Merge branch 'dev' into fix/typos

.github/workflows/publish-platform.yml

@@ -0,0 +1,108 @@
+name: publish-platform
+run-name: "platform packages ${{ inputs.version }}"
+
+on:
+  workflow_call:
+    inputs:
+      version:
+        required: true
+        type: string
+      dist_tag:
+        required: false
+        type: string
+        default: ""
+  workflow_dispatch:
+    inputs:
+      version:
+        description: "Version to publish (e.g., 3.0.0-beta.12)"
+        required: true
+        type: string
+      dist_tag:
+        description: "npm dist tag (e.g., beta, latest)"
+        required: false
+        type: string
+        default: ""
+
+permissions:
+  contents: read
+  id-token: write
+
+jobs:
+  publish-platform:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      max-parallel: 2
+      matrix:
+        platform: [darwin-arm64, darwin-x64, linux-x64, linux-arm64, linux-x64-musl, linux-arm64-musl, windows-x64]
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: oven-sh/setup-bun@v2
+        with:
+          bun-version: latest
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: "24"
+          registry-url: "https://registry.npmjs.org"
+
+      - name: Install dependencies
+        run: bun install
+        env:
+          BUN_INSTALL_ALLOW_SCRIPTS: "@ast-grep/napi"
+
+      - name: Check if already published
+        id: check
+        run: |
+          PKG_NAME="oh-my-opencode-${{ matrix.platform }}"
+          VERSION="${{ inputs.version }}"
+          STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/${PKG_NAME}/${VERSION}")
+          if [ "$STATUS" = "200" ]; then
+            echo "skip=true" >> $GITHUB_OUTPUT
+            echo "✓ ${PKG_NAME}@${VERSION} already published"
+          else
+            echo "skip=false" >> $GITHUB_OUTPUT
+            echo "→ ${PKG_NAME}@${VERSION} needs publishing"
+          fi
+
+      - name: Update version
+        if: steps.check.outputs.skip != 'true'
+        run: |
+          VERSION="${{ inputs.version }}"
+          cd packages/${{ matrix.platform }}
+          jq --arg v "$VERSION" '.version = $v' package.json > tmp.json && mv tmp.json package.json
+
+      - name: Build binary
+        if: steps.check.outputs.skip != 'true'
+        run: |
+          PLATFORM="${{ matrix.platform }}"
+          case "$PLATFORM" in
+            darwin-arm64) TARGET="bun-darwin-arm64" ;;
+            darwin-x64) TARGET="bun-darwin-x64" ;;
+            linux-x64) TARGET="bun-linux-x64" ;;
+            linux-arm64) TARGET="bun-linux-arm64" ;;
+            linux-x64-musl) TARGET="bun-linux-x64-musl" ;;
+            linux-arm64-musl) TARGET="bun-linux-arm64-musl" ;;
+            windows-x64) TARGET="bun-windows-x64" ;;
+          esac
+          
+          if [ "$PLATFORM" = "windows-x64" ]; then
+            OUTPUT="packages/${PLATFORM}/bin/oh-my-opencode.exe"
+          else
+            OUTPUT="packages/${PLATFORM}/bin/oh-my-opencode"
+          fi
+          
+          bun build src/cli/index.ts --compile --minify --target=$TARGET --outfile=$OUTPUT
+
+      - name: Publish ${{ matrix.platform }}
+        if: steps.check.outputs.skip != 'true'
+        run: |
+          cd packages/${{ matrix.platform }}
+          TAG_ARG=""
+          if [ -n "${{ inputs.dist_tag }}" ]; then
+            TAG_ARG="--tag ${{ inputs.dist_tag }}"
+          fi
+          npm publish --access public $TAG_ARG
+        env:
+          NPM_CONFIG_PROVENANCE: false

.github/workflows/publish.yml

@@ -1,5 +1,5 @@
 name: publish
-run-name: "${{ format('release {0}', inputs.bump) }}"
+run-name: "${{ format('release {0}', inputs.version || inputs.bump) }}"
 
 on:
   workflow_dispatch:
@@ -14,11 +14,11 @@ on:
           - minor
           - major
       version:
-        description: "Override version (e.g., 3.0.0-beta.6 for beta release). Takes precedence over bump."
+        description: "Override version (e.g., 3.0.0-beta.6). Takes precedence over bump."
         required: false
         type: string
       skip_platform:
-        description: "Skip platform binary packages (use when already published)"
+        description: "Skip platform binary packages"
         required: false
         type: boolean
         default: false
@@ -28,6 +28,7 @@ concurrency: ${{ github.workflow }}-${{ github.ref }}
 permissions:
   contents: write
   id-token: write
+  actions: write
 
 jobs:
   test:
@@ -64,10 +65,13 @@ jobs:
       - name: Type check
         run: bun run typecheck
 
-  publish:
+  publish-main:
     runs-on: ubuntu-latest
     needs: [test, typecheck]
     if: github.repository == 'code-yeongyu/oh-my-opencode'
+    outputs:
+      version: ${{ steps.version.outputs.version }}
+      dist_tag: ${{ steps.version.outputs.dist_tag }}
     steps:
       - uses: actions/checkout@v4
         with:
@@ -84,72 +88,152 @@ jobs:
           node-version: "24"
           registry-url: "https://registry.npmjs.org"
 
-      - name: Upgrade npm for OIDC trusted publishing
-        run: npm install -g npm@latest
-
-      - name: Configure npm registry
-        run: npm config set registry https://registry.npmjs.org
-
       - name: Install dependencies
         run: bun install
         env:
           BUN_INSTALL_ALLOW_SCRIPTS: "@ast-grep/napi"
 
-      - name: Debug environment
+      - name: Calculate version
+        id: version
         run: |
-          echo "=== Bun version ==="
+          VERSION="${{ inputs.version }}"
-          bun --version
+          if [ -z "$VERSION" ]; then
-          echo "=== Node version ==="
+            PREV=$(curl -s https://registry.npmjs.org/oh-my-opencode/latest | jq -r '.version // "0.0.0"')
-          node --version
+            BASE="${PREV%%-*}"
-          echo "=== Current directory ==="
+            IFS='.' read -r MAJOR MINOR PATCH <<< "$BASE"
-          pwd
+            case "${{ inputs.bump }}" in
-          echo "=== List src/ ==="
+              major) VERSION="$((MAJOR+1)).0.0" ;;
-          ls -la src/
+              minor) VERSION="${MAJOR}.$((MINOR+1)).0" ;;
-          echo "=== package.json scripts ==="
+              *) VERSION="${MAJOR}.${MINOR}.$((PATCH+1))" ;;
-          cat package.json | jq '.scripts'
+            esac
+          fi
+          echo "version=$VERSION" >> $GITHUB_OUTPUT
           
-      - name: Build
+          if [[ "$VERSION" == *"-"* ]]; then
+            DIST_TAG=$(echo "$VERSION" | cut -d'-' -f2 | cut -d'.' -f1)
+            echo "dist_tag=${DIST_TAG:-next}" >> $GITHUB_OUTPUT
+          else
+            echo "dist_tag=" >> $GITHUB_OUTPUT
+          fi
+          
+          echo "Version: $VERSION"
+
+      - name: Check if already published
+        id: check
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          STATUS=$(curl -s -o /dev/null -w "%{http_code}" "https://registry.npmjs.org/oh-my-opencode/${VERSION}")
+          if [ "$STATUS" = "200" ]; then
+            echo "skip=true" >> $GITHUB_OUTPUT
+            echo "✓ oh-my-opencode@${VERSION} already published"
+          else
+            echo "skip=false" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Update version
+        if: steps.check.outputs.skip != 'true'
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          jq --arg v "$VERSION" '.version = $v' package.json > tmp.json && mv tmp.json package.json
+          
+          for platform in darwin-arm64 darwin-x64 linux-x64 linux-arm64 linux-x64-musl linux-arm64-musl windows-x64; do
+            jq --arg v "$VERSION" '.version = $v' "packages/${platform}/package.json" > tmp.json
+            mv tmp.json "packages/${platform}/package.json"
+          done
+          
+          jq --arg v "$VERSION" '.optionalDependencies = (.optionalDependencies | to_entries | map(.value = $v) | from_entries)' package.json > tmp.json && mv tmp.json package.json
+
+      - name: Build main package
+        if: steps.check.outputs.skip != 'true'
         run: |
-          echo "=== Running bun build (main) ==="
           bun build src/index.ts --outdir dist --target bun --format esm --external @ast-grep/napi
-          echo "=== Running bun build (CLI) ==="
           bun build src/cli/index.ts --outdir dist/cli --target bun --format esm --external @ast-grep/napi
-          echo "=== Running tsc ==="
           bunx tsc --emitDeclarationOnly
-          echo "=== Running build:schema ==="
           bun run build:schema
 
-      - name: Build platform binaries
+      - name: Publish main package
-        run: bun run build:binaries
+        if: steps.check.outputs.skip != 'true'
-      
-      - name: Verify build output
         run: |
-          echo "=== dist/ contents ==="
+          TAG_ARG=""
-          ls -la dist/
+          if [ -n "${{ steps.version.outputs.dist_tag }}" ]; then
-          echo "=== dist/cli/ contents ==="
+            TAG_ARG="--tag ${{ steps.version.outputs.dist_tag }}"
-          ls -la dist/cli/
+          fi
-          test -f dist/index.js || (echo "ERROR: dist/index.js not found!" && exit 1)
+          npm publish --access public --provenance $TAG_ARG
-          test -f dist/cli/index.js || (echo "ERROR: dist/cli/index.js not found!" && exit 1)
-          echo "=== Platform binaries ==="
-          for platform in darwin-arm64 darwin-x64 linux-x64 linux-arm64 linux-x64-musl linux-arm64-musl; do
-            test -f "packages/${platform}/bin/oh-my-opencode" || (echo "ERROR: packages/${platform}/bin/oh-my-opencode not found!" && exit 1)
-            echo "✓ packages/${platform}/bin/oh-my-opencode"
-          done
-          test -f "packages/windows-x64/bin/oh-my-opencode.exe" || (echo "ERROR: packages/windows-x64/bin/oh-my-opencode.exe not found!" && exit 1)
-          echo "✓ packages/windows-x64/bin/oh-my-opencode.exe"
-
-      - name: Publish
-        run: bun run script/publish.ts
         env:
-          BUMP: ${{ inputs.bump }}
-          VERSION: ${{ inputs.version }}
-          SKIP_PLATFORM_PACKAGES: ${{ inputs.skip_platform }}
-          CI: true
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           NPM_CONFIG_PROVENANCE: true
 
+      - name: Git commit and tag
+        if: steps.check.outputs.skip != 'true'
+        run: |
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+          git config user.name "github-actions[bot]"
+          git add package.json assets/oh-my-opencode.schema.json packages/*/package.json || true
+          git diff --cached --quiet || git commit -m "release: v${{ steps.version.outputs.version }}"
+          git tag -f "v${{ steps.version.outputs.version }}"
+          git push origin --tags --force
+          git push origin HEAD || echo "Branch push failed (non-critical)"
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+  trigger-platform:
+    runs-on: ubuntu-latest
+    needs: publish-main
+    if: inputs.skip_platform != true
+    steps:
+      - name: Trigger platform publish workflow
+        run: |
+          gh workflow run publish-platform.yml \
+            --repo ${{ github.repository }} \
+            --ref ${{ github.ref }} \
+            -f version=${{ needs.publish-main.outputs.version }} \
+            -f dist_tag=${{ needs.publish-main.outputs.dist_tag }}
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+  release:
+    runs-on: ubuntu-latest
+    needs: publish-main
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Generate changelog
+        id: changelog
+        run: |
+          VERSION="${{ needs.publish-main.outputs.version }}"
+          
+          PREV_TAG=""
+          if [[ "$VERSION" == *"-beta."* ]]; then
+            BASE="${VERSION%-beta.*}"
+            NUM="${VERSION##*-beta.}"
+            PREV_NUM=$((NUM - 1))
+            if [ $PREV_NUM -ge 1 ]; then
+              PREV_TAG="${BASE}-beta.${PREV_NUM}"
+              git rev-parse "v${PREV_TAG}" >/dev/null 2>&1 || PREV_TAG=""
+            fi
+          fi
+          
+          if [ -z "$PREV_TAG" ]; then
+            PREV_TAG=$(curl -s https://registry.npmjs.org/oh-my-opencode/latest | jq -r '.version // "0.0.0"')
+          fi
+          
+          echo "Comparing v${PREV_TAG}..v${VERSION}"
+          
+          NOTES=$(git log "v${PREV_TAG}..v${VERSION}" --oneline --format="- %h %s" 2>/dev/null | grep -vE "^- \w+ (ignore:|test:|chore:|ci:|release:)" || echo "No notable changes")
+          
+          echo "$NOTES" > /tmp/changelog.md
+
+      - name: Create GitHub release
+        run: |
+          VERSION="${{ needs.publish-main.outputs.version }}"
+          gh release view "v${VERSION}" >/dev/null 2>&1 || \
+            gh release create "v${VERSION}" --title "v${VERSION}" --notes-file /tmp/changelog.md
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
       - name: Delete draft release
-        run: gh release delete next --yes 2>/dev/null || echo "No draft release to delete"
+        run: gh release delete next --yes 2>/dev/null || true
         env:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
@@ -158,8 +242,10 @@ jobs:
         run: |
           git config user.name "github-actions[bot]"
           git config user.email "github-actions[bot]@users.noreply.github.com"
-          VERSION=$(jq -r '.version' package.json)
+          VERSION="${{ needs.publish-main.outputs.version }}"
           git stash --include-untracked || true
           git checkout master
           git reset --hard "v${VERSION}"
-          git push -f origin master || echo "::warning::Failed to push to master. This can happen when workflow files changed. Manually sync master: git checkout master && git reset --hard v${VERSION} && git push -f"
+          git push -f origin master || echo "::warning::Failed to push to master"
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/sisyphus-agent.yml

@@ -103,7 +103,7 @@ jobs:
           opencode --version
 
           # Run local oh-my-opencode install (uses built dist)
-          bun run dist/cli/index.js install --no-tui --claude=max20 --chatgpt=no --gemini=no --copilot=no
+          bun run dist/cli/index.js install --no-tui --claude=max20 --openai=no --gemini=no --copilot=no
 
           # Override plugin to use local file reference
           OPENCODE_JSON=~/.config/opencode/opencode.json
@@ -114,6 +114,7 @@ jobs:
 
           OPENCODE_JSON=~/.config/opencode/opencode.json
           jq --arg baseURL "$ANTHROPIC_BASE_URL" --arg apiKey "$ANTHROPIC_API_KEY" '
+            .model = "anthropic/claude-opus-4-5" |
             .provider.anthropic = {
               "name": "Anthropic",
               "npm": "@ai-sdk/anthropic",

AGENTS.md

@@ -1,12 +1,12 @@
 # PROJECT KNOWLEDGE BASE
 
-**Generated:** 2026-01-19T18:10:00+09:00
+**Generated:** 2026-01-20T17:18:00+09:00
-**Commit:** 45660940
+**Commit:** 3d3d3e49
 **Branch:** dev
 
 ## OVERVIEW
 
-OpenCode plugin implementing multi-model agent orchestration (Claude Opus 4.5, GPT-5.2, Gemini 3, Grok, GLM-4.7). 31 lifecycle hooks, 20+ tools (LSP, AST-Grep, delegation), 10 specialized agents, Claude Code compatibility layer. "oh-my-zsh" for OpenCode.
+ClaudeCode plugin implementing multi-model agent orchestration (Claude Opus 4.5, GPT-5.2, Gemini 3, Grok, GLM-4.7). 31 lifecycle hooks, 20+ tools (LSP, AST-Grep, delegation), 10 specialized agents, Claude Code compatibility layer. "oh-my-zsh" for ClaudeCode.
 
 ## STRUCTURE
 
@@ -21,7 +21,7 @@ oh-my-opencode/
 │   ├── cli/           # CLI installer, doctor, run - see src/cli/AGENTS.md
 │   ├── mcp/           # Built-in MCPs: websearch, context7, grep_app
 │   ├── config/        # Zod schema, TypeScript types
-│   └── index.ts       # Main plugin entry (568 lines)
+│   └── index.ts       # Main plugin entry (589 lines)
 ├── script/            # build-schema.ts, publish.ts, build-binaries.ts
 ├── packages/          # 7 platform-specific binaries
 └── dist/              # Build output (ESM + .d.ts)
@@ -44,7 +44,7 @@ oh-my-opencode/
 | Skill MCP | `src/features/skill-mcp-manager/` | MCP servers embedded in skills |
 | CLI installer | `src/cli/install.ts` | Interactive TUI (462 lines) |
 | Doctor checks | `src/cli/doctor/checks/` | 14 health checks across 6 categories |
-| Orchestrator | `src/hooks/sisyphus-orchestrator/` | Main orchestration hook (771 lines) |
+| Orchestrator | `src/hooks/atlas/` | Main orchestration hook (771 lines) |
 
 ## TDD (Test-Driven Development)
 
@@ -109,8 +109,6 @@ oh-my-opencode/
 | oracle | openai/gpt-5.2 | Read-only consultation, high-IQ debugging |
 | librarian | opencode/glm-4.7-free | Multi-repo analysis, docs, GitHub search |
 | explore | opencode/grok-code | Fast codebase exploration (contextual grep) |
-| frontend-ui-ux-engineer | google/gemini-3-pro-preview | UI generation, visual design |
-| document-writer | google/gemini-3-flash | Technical documentation |
 | multimodal-looker | google/gemini-3-flash | PDF/image analysis |
 | Prometheus (Planner) | anthropic/claude-opus-4-5 | Strategic planning, interview mode |
 | Metis (Plan Consultant) | anthropic/claude-sonnet-4-5 | Pre-planning analysis |
@@ -145,14 +143,14 @@ bun test               # Run tests (83 test files)
 
 | File | Lines | Description |
 |------|-------|-------------|
-| `src/agents/orchestrator-sisyphus.ts` | 1531 | Orchestrator agent, 7-section delegation, wisdom accumulation |
+| `src/agents/atlas.ts` | 1383 | Orchestrator agent, 7-section delegation, wisdom accumulation |
 | `src/features/builtin-skills/skills.ts` | 1203 | Skill definitions (playwright, git-master, frontend-ui-ux) |
 | `src/agents/prometheus-prompt.ts` | 1196 | Planning agent, interview mode, Momus loop |
 | `src/features/background-agent/manager.ts` | 1165 | Task lifecycle, concurrency, notification batching |
-| `src/hooks/sisyphus-orchestrator/index.ts` | 771 | Orchestrator hook implementation |
+| `src/hooks/atlas/index.ts` | 771 | Orchestrator hook implementation |
-| `src/tools/delegate-task/tools.ts` | 761 | Category-based task delegation |
+| `src/tools/delegate-task/tools.ts` | 770 | Category-based task delegation |
-| `src/cli/config-manager.ts` | 730 | JSONC parsing, multi-level config |
+| `src/cli/config-manager.ts` | 616 | JSONC parsing, multi-level config |
-| `src/agents/sisyphus.ts` | 640 | Main Sisyphus prompt |
+| `src/agents/sisyphus.ts` | 615 | Main Sisyphus prompt |
 | `src/features/builtin-commands/templates/refactor.ts` | 619 | Refactoring command template |
 | `src/tools/lsp/client.ts` | 596 | LSP protocol, JSON-RPC |
 
@@ -173,7 +171,7 @@ Three-tier MCP system:
 ## NOTES
 
 - **Testing**: Bun native test (`bun test`), BDD-style, 83 test files
-- **OpenCode**: Requires >= 1.0.150
+- **ClaudeCode**: Requires >= 1.0.150
 - **Multi-lang docs**: README.md (EN), README.ko.md (KO), README.ja.md (JA), README.zh-cn.md (ZH-CN)
 - **Config**: `~/.config/opencode/oh-my-opencode.json` (user) or `.opencode/oh-my-opencode.json` (project)
 - **Trusted deps**: @ast-grep/cli, @ast-grep/napi, @code-yeongyu/comment-checker

README.ja.md

@@ -16,7 +16,7 @@
 
 > [!TIP]
 >
-> [![The Orchestrator is now available in beta.](./.github/assets/orchestrator-sisyphus.png?v=3)](https://github.com/code-yeongyu/oh-my-opencode/releases/tag/v3.0.0-beta.10)
+> [![The Orchestrator is now available in beta.](./.github/assets/orchestrator-atlas.png?v=3)](https://github.com/code-yeongyu/oh-my-opencode/releases/tag/v3.0.0-beta.10)
 > > **オーケストレーターがベータ版で利用可能になりました。`oh-my-opencode@3.0.0-beta.10`を使用してインストールしてください。**
 >
 > 一緒に歩みましょう！
@@ -279,7 +279,7 @@ oh-my-opencode を削除するには：
 詳細は [Features Documentation](docs/features.md) を参照してください。
 
 **概要:**
-- **エージェント**: Sisyphus（メインエージェント）、Prometheus（プランナー）、Oracle（アーキテクチャ/デバッグ）、Librarian（ドキュメント/コード検索）、Explore（高速コードベース grep）、Frontend Engineer（UI/UX）、Document Writer、Multimodal Looker
+- **エージェント**: Sisyphus（メインエージェント）、Prometheus（プランナー）、Oracle（アーキテクチャ/デバッグ）、Librarian（ドキュメント/コード検索）、Explore（高速コードベース grep）、Multimodal Looker
 - **バックグラウンドエージェント**: 本物の開発チームのように複数エージェントを並列実行
 - **LSP & AST ツール**: リファクタリング、リネーム、診断、AST 認識コード検索
 - **コンテキスト注入**: AGENTS.md、README.md、条件付きルールの自動注入

README.md

@@ -16,7 +16,7 @@
 
 > [!TIP]
 >
-> [![The Orchestrator is now available in beta.](./.github/assets/orchestrator-sisyphus.png?v=3)](https://github.com/code-yeongyu/oh-my-opencode/releases/tag/v3.0.0-beta.10)
+> [![The Orchestrator is now available in beta.](./.github/assets/orchestrator-atlas.png?v=3)](https://github.com/code-yeongyu/oh-my-opencode/releases/tag/v3.0.0-beta.10)
 > > **The Orchestrator is now available in beta. Use `oh-my-opencode@3.0.0-beta.10` to install it.**
 >
 > Be with us!
@@ -293,7 +293,7 @@ We have lots of features that you'll think should obviously exist, and once you
 See the full [Features Documentation](docs/features.md) for detailed information.
 
 **Quick Overview:**
-- **Agents**: Sisyphus (the main agent), Prometheus (planner), Oracle (architecture/debugging), Librarian (docs/code search), Explore (fast codebase grep), Frontend Engineer (UI/UX), Document Writer, Multimodal Looker
+- **Agents**: Sisyphus (the main agent), Prometheus (planner), Oracle (architecture/debugging), Librarian (docs/code search), Explore (fast codebase grep), Multimodal Looker
 - **Background Agents**: Run multiple agents in parallel like a real dev team
 - **LSP & AST Tools**: Refactoring, rename, diagnostics, AST-aware code search
 - **Context Injection**: Auto-inject AGENTS.md, README.md, conditional rules

README.zh-cn.md

@@ -16,7 +16,7 @@
 
 > [!TIP]
 >
-> [![Orchestrator 现已进入测试阶段。](./.github/assets/orchestrator-sisyphus.png?v=3)](https://github.com/code-yeongyu/oh-my-opencode/releases/tag/v3.0.0-beta.10)
+> [![Orchestrator 现已进入测试阶段。](./.github/assets/orchestrator-atlas.png?v=3)](https://github.com/code-yeongyu/oh-my-opencode/releases/tag/v3.0.0-beta.10)
 > > **Orchestrator 现已进入测试阶段。使用 `oh-my-opencode@3.0.0-beta.10` 安装。**
 >
 > 加入我们！
@@ -289,7 +289,7 @@ curl -s https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/refs/heads
 详细信息请参阅 [Features Documentation](docs/features.md)。
 
 **概览：**
-- **智能体**：Sisyphus（主智能体）、Prometheus（规划器）、Oracle（架构/调试）、Librarian（文档/代码搜索）、Explore（快速代码库 grep）、Frontend Engineer（UI/UX）、Document Writer、Multimodal Looker
+- **智能体**：Sisyphus（主智能体）、Prometheus（规划器）、Oracle（架构/调试）、Librarian（文档/代码搜索）、Explore（快速代码库 grep）、Multimodal Looker
 - **后台智能体**：像真正的开发团队一样并行运行多个智能体
 - **LSP & AST 工具**：重构、重命名、诊断、AST 感知代码搜索
 - **上下文注入**：自动注入 AGENTS.md、README.md、条件规则

assets/oh-my-opencode.schema.json

@@ -24,12 +24,10 @@
           "oracle",
           "librarian",
           "explore",
-          "frontend-ui-ux-engineer",
-          "document-writer",
           "multimodal-looker",
           "Metis (Plan Consultant)",
           "Momus (Plan Reviewer)",
-          "orchestrator-sisyphus"
+          "Atlas"
         ]
       }
     },
@@ -78,7 +76,7 @@
           "delegate-task-retry",
           "prometheus-md-only",
           "start-work",
-          "sisyphus-orchestrator"
+          "atlas"
         ]
       }
     },
@@ -1481,258 +1479,6 @@
             }
           }
         },
-        "frontend-ui-ux-engineer": {
-          "type": "object",
-          "properties": {
-            "model": {
-              "type": "string"
-            },
-            "variant": {
-              "type": "string"
-            },
-            "category": {
-              "type": "string"
-            },
-            "skills": {
-              "type": "array",
-              "items": {
-                "type": "string"
-              }
-            },
-            "temperature": {
-              "type": "number",
-              "minimum": 0,
-              "maximum": 2
-            },
-            "top_p": {
-              "type": "number",
-              "minimum": 0,
-              "maximum": 1
-            },
-            "prompt": {
-              "type": "string"
-            },
-            "prompt_append": {
-              "type": "string"
-            },
-            "tools": {
-              "type": "object",
-              "propertyNames": {
-                "type": "string"
-              },
-              "additionalProperties": {
-                "type": "boolean"
-              }
-            },
-            "disable": {
-              "type": "boolean"
-            },
-            "description": {
-              "type": "string"
-            },
-            "mode": {
-              "type": "string",
-              "enum": [
-                "subagent",
-                "primary",
-                "all"
-              ]
-            },
-            "color": {
-              "type": "string",
-              "pattern": "^#[0-9A-Fa-f]{6}$"
-            },
-            "permission": {
-              "type": "object",
-              "properties": {
-                "edit": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                },
-                "bash": {
-                  "anyOf": [
-                    {
-                      "type": "string",
-                      "enum": [
-                        "ask",
-                        "allow",
-                        "deny"
-                      ]
-                    },
-                    {
-                      "type": "object",
-                      "propertyNames": {
-                        "type": "string"
-                      },
-                      "additionalProperties": {
-                        "type": "string",
-                        "enum": [
-                          "ask",
-                          "allow",
-                          "deny"
-                        ]
-                      }
-                    }
-                  ]
-                },
-                "webfetch": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                },
-                "doom_loop": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                },
-                "external_directory": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                }
-              }
-            }
-          }
-        },
-        "document-writer": {
-          "type": "object",
-          "properties": {
-            "model": {
-              "type": "string"
-            },
-            "variant": {
-              "type": "string"
-            },
-            "category": {
-              "type": "string"
-            },
-            "skills": {
-              "type": "array",
-              "items": {
-                "type": "string"
-              }
-            },
-            "temperature": {
-              "type": "number",
-              "minimum": 0,
-              "maximum": 2
-            },
-            "top_p": {
-              "type": "number",
-              "minimum": 0,
-              "maximum": 1
-            },
-            "prompt": {
-              "type": "string"
-            },
-            "prompt_append": {
-              "type": "string"
-            },
-            "tools": {
-              "type": "object",
-              "propertyNames": {
-                "type": "string"
-              },
-              "additionalProperties": {
-                "type": "boolean"
-              }
-            },
-            "disable": {
-              "type": "boolean"
-            },
-            "description": {
-              "type": "string"
-            },
-            "mode": {
-              "type": "string",
-              "enum": [
-                "subagent",
-                "primary",
-                "all"
-              ]
-            },
-            "color": {
-              "type": "string",
-              "pattern": "^#[0-9A-Fa-f]{6}$"
-            },
-            "permission": {
-              "type": "object",
-              "properties": {
-                "edit": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                },
-                "bash": {
-                  "anyOf": [
-                    {
-                      "type": "string",
-                      "enum": [
-                        "ask",
-                        "allow",
-                        "deny"
-                      ]
-                    },
-                    {
-                      "type": "object",
-                      "propertyNames": {
-                        "type": "string"
-                      },
-                      "additionalProperties": {
-                        "type": "string",
-                        "enum": [
-                          "ask",
-                          "allow",
-                          "deny"
-                        ]
-                      }
-                    }
-                  ]
-                },
-                "webfetch": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                },
-                "doom_loop": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                },
-                "external_directory": {
-                  "type": "string",
-                  "enum": [
-                    "ask",
-                    "allow",
-                    "deny"
-                  ]
-                }
-              }
-            }
-          }
-        },
         "multimodal-looker": {
           "type": "object",
           "properties": {
@@ -1859,7 +1605,7 @@
             }
           }
         },
-        "orchestrator-sisyphus": {
+        "Atlas": {
           "type": "object",
           "properties": {
             "model": {
@@ -2059,6 +1805,9 @@
           },
           "prompt_append": {
             "type": "string"
+          },
+          "is_unstable_agent": {
+            "type": "boolean"
           }
         }
       }

bun.lock

@@ -27,13 +27,13 @@
         "typescript": "^5.7.3",
       },
       "optionalDependencies": {
-        "oh-my-opencode-darwin-arm64": "3.0.0-beta.8",
+        "oh-my-opencode-darwin-arm64": "3.0.0-beta.11",
-        "oh-my-opencode-darwin-x64": "3.0.0-beta.8",
+        "oh-my-opencode-darwin-x64": "3.0.0-beta.11",
-        "oh-my-opencode-linux-arm64": "3.0.0-beta.8",
+        "oh-my-opencode-linux-arm64": "3.0.0-beta.11",
-        "oh-my-opencode-linux-arm64-musl": "3.0.0-beta.8",
+        "oh-my-opencode-linux-arm64-musl": "3.0.0-beta.11",
-        "oh-my-opencode-linux-x64": "3.0.0-beta.8",
+        "oh-my-opencode-linux-x64": "3.0.0-beta.11",
-        "oh-my-opencode-linux-x64-musl": "3.0.0-beta.8",
+        "oh-my-opencode-linux-x64-musl": "3.0.0-beta.11",
-        "oh-my-opencode-windows-x64": "3.0.0-beta.8",
+        "oh-my-opencode-windows-x64": "3.0.0-beta.11",
       },
     },
   },
@@ -225,6 +225,20 @@
 
     "object-inspect": ["object-inspect@1.13.4", "", {}, "sha512-W67iLl4J2EXEGTbfeHCffrjDfitvLANg0UlX3wFUUSTx92KXRFegMHUVgSqE+wvhAbi4WqjGg9czysTV2Epbew=="],
 
+    "oh-my-opencode-darwin-arm64": ["oh-my-opencode-darwin-arm64@3.0.0-beta.11", "", { "os": "darwin", "cpu": "arm64", "bin": { "oh-my-opencode": "bin/oh-my-opencode" } }, "sha512-7cFv2bbz9HTY7sshgVTu+IhvYf7CT0czDYqHEB+dYfEqFU6TaoSMimq6uHqcWegUUR1T7PNmc0dyjYVw69FeVA=="],
+
+    "oh-my-opencode-darwin-x64": ["oh-my-opencode-darwin-x64@3.0.0-beta.11", "", { "os": "darwin", "cpu": "x64", "bin": { "oh-my-opencode": "bin/oh-my-opencode" } }, "sha512-rGAbDdUySWITIdm2yiuNFB9lFYaSXT8LMtg97LTlOO5vZbI3M+obIS3QlIkBtAhgOTIPB7Ni+T0W44OmJpHoYA=="],
+
+    "oh-my-opencode-linux-arm64": ["oh-my-opencode-linux-arm64@3.0.0-beta.11", "", { "os": "linux", "cpu": "arm64", "bin": { "oh-my-opencode": "bin/oh-my-opencode" } }, "sha512-F9dqwWwGAdqeSkE7Tre5DmHQXwDpU2Z8Jk0lwTJMLj+kMqYFDVPjLPo4iVUdwPpxpmm0pR84u/oonG/2+84/zw=="],
+
+    "oh-my-opencode-linux-arm64-musl": ["oh-my-opencode-linux-arm64-musl@3.0.0-beta.11", "", { "os": "linux", "cpu": "arm64", "bin": { "oh-my-opencode": "bin/oh-my-opencode" } }, "sha512-H+zOtHkHd+TmdPj64M1A0zLOk7OHIK4C8yqfLFhfizOIBffT1yOhAs6EpK3EqPhfPLu54ADgcQcu8W96VP24UA=="],
+
+    "oh-my-opencode-linux-x64": ["oh-my-opencode-linux-x64@3.0.0-beta.11", "", { "os": "linux", "cpu": "x64", "bin": { "oh-my-opencode": "bin/oh-my-opencode" } }, "sha512-IG+KODTJ8rs6cEJ2wN6Zpr6YtvCS5OpYP6jBdGJltmUpjQdMhdMsaY3ysZk+9Vxpx2KC3xj5KLHV1USg3uBTeg=="],
+
+    "oh-my-opencode-linux-x64-musl": ["oh-my-opencode-linux-x64-musl@3.0.0-beta.11", "", { "os": "linux", "cpu": "x64", "bin": { "oh-my-opencode": "bin/oh-my-opencode" } }, "sha512-irV+AuWrHqNm7VT7HO56qgymR0+vEfJbtB3vCq68kprH2V4NQmGp2MNKIYPnUCYL7NEK3H2NX+h06YFZJ/8ELQ=="],
+
+    "oh-my-opencode-windows-x64": ["oh-my-opencode-windows-x64@3.0.0-beta.11", "", { "os": "win32", "cpu": "x64", "bin": { "oh-my-opencode": "bin/oh-my-opencode.exe" } }, "sha512-exZ/NEwGBlxyWszN7dvOfzbYX0cuhBZXftqAAFOlVP26elDHdo+AmSmLR/4cJyzpR9nCWz4xvl/RYF84bY6OEA=="],
+
     "on-finished": ["on-finished@2.4.1", "", { "dependencies": { "ee-first": "1.1.1" } }, "sha512-oVlzkg3ENAhCk2zdv7IJwd/QUD4z2RxRwpkcGY8psCVcCYZNq4wYnVWALHM+brtuJjePWiYF/ClmuDr8Ch5+kg=="],
 
     "once": ["once@1.4.0", "", { "dependencies": { "wrappy": "1" } }, "sha512-lNaJgI+2Q5URQBkccEKHTQOPaXdUxnZZElQTZY0MFUAuaEqe1E+Nyvgdz/aIyNi6Z9MzO5dv1H8n58/GELp3+w=="],

docs/configurations.md

@@ -63,7 +63,7 @@ Override built-in agent settings:
       "model": "anthropic/claude-haiku-4-5",
       "temperature": 0.5
     },
-    "frontend-ui-ux-engineer": {
+    "multimodal-looker": {
       "disable": true
     }
   }
@@ -116,11 +116,11 @@ Or disable via `disabled_agents` in `~/.config/opencode/oh-my-opencode.json` or
 
 ```json
 {
-  "disabled_agents": ["oracle", "frontend-ui-ux-engineer"]
+  "disabled_agents": ["oracle", "multimodal-looker"]
 }
 ```
 
-Available agents: `oracle`, `librarian`, `explore`, `frontend-ui-ux-engineer`, `document-writer`, `multimodal-looker`
+Available agents: `oracle`, `librarian`, `explore`, `multimodal-looker`
 
 ## Built-in Skills
 
@@ -308,6 +308,268 @@ Add custom categories in `oh-my-opencode.json`:
 
 Each category supports: `model`, `temperature`, `top_p`, `maxTokens`, `thinking`, `reasoningEffort`, `textVerbosity`, `tools`, `prompt_append`.
 
+## Model Selection System
+
+The installer automatically configures optimal models based on your subscriptions. This section explains how models are selected for each agent and category.
+
+### Overview
+
+**Problem**: Users have different subscription combinations (Claude, OpenAI, Gemini, etc.). The system needs to automatically select the best available model for each task.
+
+**Solution**: A tiered fallback system that:
+1. Prioritizes native provider subscriptions (Claude, OpenAI, Gemini)
+2. Falls back through alternative providers in priority order
+3. Applies capability-specific logic (e.g., Oracle prefers GPT, visual tasks prefer Gemini)
+
+### Provider Priority
+
+```
+┌─────────────────────────────────────────────────────────────────┐
+│                     MODEL SELECTION FLOW                        │
+├─────────────────────────────────────────────────────────────────┤
+│                                                                 │
+│   ┌─────────────────────────────────────────────────────────┐   │
+│   │              TIER 1: NATIVE PROVIDERS                   │   │
+│   │         (Your direct subscriptions)                     │   │
+│   │                                                         │   │
+│   │   Claude (anthropic/) ──► OpenAI (openai/) ──► Gemini   │   │
+│   │         │                      │              (google/) │   │
+│   │         ▼                      ▼                   │    │   │
+│   │   Opus/Sonnet/Haiku    GPT-5.2/Codex      Gemini 3 Pro │   │
+│   └─────────────────────────────────────────────────────────┘   │
+│                              │                                  │
+│                              ▼ (if no native available)         │
+│   ┌─────────────────────────────────────────────────────────┐   │
+│   │              TIER 2: OPENCODE ZEN                       │   │
+│   │         (opencode/ prefix models)                       │   │
+│   │                                                         │   │
+│   │   opencode/claude-opus-4-5, opencode/gpt-5.2, etc.      │   │
+│   └─────────────────────────────────────────────────────────┘   │
+│                              │                                  │
+│                              ▼ (if no OpenCode Zen)             │
+│   ┌─────────────────────────────────────────────────────────┐   │
+│   │              TIER 3: GITHUB COPILOT                     │   │
+│   │         (github-copilot/ prefix models)                 │   │
+│   │                                                         │   │
+│   │   github-copilot/claude-opus-4.5, etc.                  │   │
+│   └─────────────────────────────────────────────────────────┘   │
+│                              │                                  │
+│                              ▼ (if no Copilot)                  │
+│   ┌─────────────────────────────────────────────────────────┐   │
+│   │              TIER 4: Z.AI CODING PLAN                   │   │
+│   │         (zai-coding-plan/ prefix models)                │   │
+│   │                                                         │   │
+│   │   zai-coding-plan/glm-4.7 (GLM models only)             │   │
+│   └─────────────────────────────────────────────────────────┘   │
+│                              │                                  │
+│                              ▼ (ultimate fallback)              │
+│   ┌─────────────────────────────────────────────────────────┐   │
+│   │              FALLBACK: FREE TIER                        │   │
+│   │                                                         │   │
+│   │   opencode/glm-4.7-free                                 │   │
+│   └─────────────────────────────────────────────────────────┘   │
+│                                                                 │
+└─────────────────────────────────────────────────────────────────┘
+```
+
+### Native Tier Cross-Fallback
+
+Within the Native tier, models fall back based on capability requirements:
+
+| Capability | 1st Choice | 2nd Choice | 3rd Choice |
+|------------|------------|------------|------------|
+| **High-tier tasks** (Sisyphus, Atlas) | Claude Opus | OpenAI GPT-5.2 | Gemini 3 Pro |
+| **Standard tasks** | Claude Sonnet | OpenAI GPT-5.2 | Gemini 3 Flash |
+| **Quick tasks** | Claude Haiku | OpenAI GPT-5.1-mini | Gemini 3 Flash |
+| **Deep reasoning** (Oracle) | OpenAI GPT-5.2-Codex | Claude Opus | Gemini 3 Pro |
+| **Visual/UI tasks** | Gemini 3 Pro | OpenAI GPT-5.2 | Claude Sonnet |
+| **Writing tasks** | Gemini 3 Flash | OpenAI GPT-5.2 | Claude Sonnet |
+
+### Agent-Specific Rules
+
+#### Standard Agents
+
+| Agent | Capability | Example (Claude + OpenAI + Gemini) |
+|-------|------------|-------------------------------------|
+| **Sisyphus** | High-tier (isMax20) or Standard | `anthropic/claude-opus-4-5` or `anthropic/claude-sonnet-4-5` |
+| **Oracle** | Deep reasoning | `openai/gpt-5.2-codex` |
+| **Prometheus** | High-tier/Standard | Same as Sisyphus |
+| **Metis** | High-tier/Standard | Same as Sisyphus |
+| **Momus** | Deep reasoning | `openai/gpt-5.2-codex` |
+| **Atlas** | High-tier/Standard | Same as Sisyphus |
+| **multimodal-looker** | Visual | `google/gemini-3-pro-preview` |
+
+#### Special Case: explore Agent
+
+The `explore` agent has unique logic for cost optimization:
+
+```
+┌────────────────────────────────────────┐
+│           EXPLORE AGENT LOGIC          │
+├────────────────────────────────────────┤
+│                                        │
+│   Has Claude + isMax20?                │
+│         │                              │
+│    YES  │  NO                          │
+│    ▼    │  ▼                           │
+│ ┌──────┐│┌────────────────────┐        │
+│ │Haiku ││ │ opencode/grok-code │        │
+│ │4.5   │││ (free & fast)       │        │
+│ └──────┘│└────────────────────┘        │
+│                                        │
+│ Rationale:                             │
+│ • max20 users want to use Claude quota │
+│ • Others save quota with free grok     │
+└────────────────────────────────────────┘
+```
+
+#### Special Case: librarian Agent
+
+The `librarian` agent prioritizes Z.ai when available:
+
+```
+┌────────────────────────────────────────┐
+│          LIBRARIAN AGENT LOGIC         │
+├────────────────────────────────────────┤
+│                                        │
+│   Has Z.ai Coding Plan?                │
+│         │                              │
+│    YES  │  NO                          │
+│    ▼    │  ▼                           │
+│ ┌──────────────┐ ┌──────────────────┐  │
+│ │zai-coding-   │ │ Normal fallback  │  │
+│ │plan/glm-4.7  │ │ chain applies    │  │
+│ └──────────────┘ └──────────────────┘  │
+│                                        │
+│ Rationale:                             │
+│ • GLM excels at documentation tasks    │
+│ • Z.ai provides dedicated GLM access   │
+└────────────────────────────────────────┘
+```
+
+### Category-Specific Rules
+
+Categories follow the same fallback logic as agents:
+
+| Category | Primary Capability | Fallback Chain |
+|----------|-------------------|----------------|
+| `visual-engineering` | Visual | Gemini → OpenAI → Claude |
+| `ultrabrain` | Deep reasoning | OpenAI → Claude → Gemini |
+| `artistry` | Visual/Creative | Gemini → OpenAI → Claude |
+| `quick` | Quick tasks | Claude Haiku → OpenAI mini → Gemini Flash |
+| `unspecified-low` | Standard | Claude Sonnet → OpenAI → Gemini Flash |
+| `unspecified-high` | High-tier | Claude Opus → OpenAI → Gemini Pro |
+| `writing` | Writing | Gemini Flash → OpenAI → Claude |
+
+### Subscription Scenarios
+
+#### Scenario 1: Claude Only (Standard Plan)
+
+```json
+// User has: Claude Pro (not max20)
+{
+  "agents": {
+    "Sisyphus": { "model": "anthropic/claude-sonnet-4-5" },
+    "oracle": { "model": "anthropic/claude-opus-4-5" },
+    "explore": { "model": "opencode/grok-code" },
+    "librarian": { "model": "opencode/glm-4.7-free" }
+  }
+}
+```
+
+#### Scenario 2: Claude Only (Max20 Plan)
+
+```json
+// User has: Claude Max (max20 mode)
+{
+  "agents": {
+    "Sisyphus": { "model": "anthropic/claude-opus-4-5" },
+    "oracle": { "model": "anthropic/claude-opus-4-5" },
+    "explore": { "model": "anthropic/claude-haiku-4-5" },
+    "librarian": { "model": "opencode/glm-4.7-free" }
+  }
+}
+```
+
+#### Scenario 3: ChatGPT Only
+
+```json
+// User has: OpenAI/ChatGPT Plus only
+{
+  "agents": {
+    "Sisyphus": { "model": "openai/gpt-5.2" },
+    "oracle": { "model": "openai/gpt-5.2-codex" },
+    "explore": { "model": "opencode/grok-code" },
+    "multimodal-looker": { "model": "openai/gpt-5.2" },
+    "librarian": { "model": "opencode/glm-4.7-free" }
+  }
+}
+```
+
+#### Scenario 4: Full Stack (Claude + OpenAI + Gemini)
+
+```json
+// User has: All native providers
+{
+  "agents": {
+    "Sisyphus": { "model": "anthropic/claude-opus-4-5" },
+    "oracle": { "model": "openai/gpt-5.2-codex" },
+    "explore": { "model": "anthropic/claude-haiku-4-5" },
+    "multimodal-looker": { "model": "google/gemini-3-pro-preview" },
+    "librarian": { "model": "opencode/glm-4.7-free" }
+  }
+}
+```
+
+#### Scenario 5: GitHub Copilot Only
+
+```json
+// User has: GitHub Copilot only (no native providers)
+{
+  "agents": {
+    "Sisyphus": { "model": "github-copilot/claude-sonnet-4.5" },
+    "oracle": { "model": "github-copilot/gpt-5.2-codex" },
+    "explore": { "model": "opencode/grok-code" },
+    "librarian": { "model": "github-copilot/gpt-5.2" }
+  }
+}
+```
+
+### isMax20 Flag Impact
+
+The `isMax20` flag (Claude Max 20x mode) affects high-tier task model selection:
+
+| isMax20 | High-tier Capability | Result |
+|---------|---------------------|--------|
+| `true` | Uses `unspecified-high` | Opus-class models |
+| `false` | Uses `unspecified-low` | Sonnet-class models |
+
+**Affected agents**: Sisyphus, Prometheus, Metis, Atlas
+
+**Why?**: Max20 users have 20x more Claude usage, so they can afford Opus for orchestration. Standard users should conserve quota with Sonnet.
+
+### Manual Override
+
+You can always override automatic selection in `oh-my-opencode.json`:
+
+```json
+{
+  "agents": {
+    "Sisyphus": {
+      "model": "anthropic/claude-sonnet-4-5"  // Force specific model
+    },
+    "oracle": {
+      "model": "openai/o3"  // Use different model
+    }
+  },
+  "categories": {
+    "visual-engineering": {
+      "model": "anthropic/claude-opus-4-5"  // Override category default
+    }
+  }
+}
+```
+
 ## Hooks
 
 Disable specific built-in hooks via `disabled_hooks` in `~/.config/opencode/oh-my-opencode.json` or `.opencode/oh-my-opencode.json`:

docs/features.md

@@ -1,161 +1,351 @@
 # Oh-My-OpenCode Features
 
-## Agents: Your Teammates
+---
 
-- **Sisyphus** (`anthropic/claude-opus-4-5`): **The default agent.** A powerful AI orchestrator for OpenCode. Plans, delegates, and executes complex tasks using specialized subagents with aggressive parallel execution. Emphasizes background task delegation and todo-driven workflow. Uses Claude Opus 4.5 with extended thinking (32k budget) for maximum reasoning capability.
+## Agents: Your AI Team
-- **oracle** (`openai/gpt-5.2`): Architecture, code review, strategy. Uses GPT-5.2 for its stellar logical reasoning and deep analysis. Inspired by AmpCode.
+
-- **librarian** (`opencode/glm-4.7-free`): Multi-repo analysis, doc lookup, implementation examples. Uses GLM-4.7 Free for deep codebase understanding and GitHub research with evidence-based answers. Inspired by AmpCode.
+Oh-My-OpenCode provides 10 specialized AI agents. Each has distinct expertise, optimized models, and tool permissions.
-- **explore** (`opencode/grok-code`, `google/gemini-3-flash`, or `anthropic/claude-haiku-4-5`): Fast codebase exploration and pattern matching. Uses Gemini 3 Flash when Antigravity auth is configured, Haiku when Claude max20 is available, otherwise Grok. Inspired by Claude Code.
+
-- **frontend-ui-ux-engineer** (`google/gemini-3-pro-preview`): A designer turned developer. Builds gorgeous UIs. Gemini excels at creative, beautiful UI code.
+### Core Agents
-- **document-writer** (`google/gemini-3-flash`): Technical writing expert. Gemini is a wordsmith—writes prose that flows.
+
-- **multimodal-looker** (`google/gemini-3-flash`): Visual content specialist. Analyzes PDFs, images, diagrams to extract information.
+| Agent | Model | Purpose |
+|-------|-------|---------|
+| **Sisyphus** | `anthropic/claude-opus-4-5` | **The default orchestrator.** Plans, delegates, and executes complex tasks using specialized subagents with aggressive parallel execution. Todo-driven workflow with extended thinking (32k budget). |
+| **oracle** | `openai/gpt-5.2` | Architecture decisions, code review, debugging. Read-only consultation - stellar logical reasoning and deep analysis. Inspired by AmpCode. |
+| **librarian** | `opencode/glm-4.7-free` | Multi-repo analysis, documentation lookup, OSS implementation examples. Deep codebase understanding with evidence-based answers. Inspired by AmpCode. |
+| **explore** | `opencode/grok-code` | Fast codebase exploration and contextual grep. Uses Gemini 3 Flash when Antigravity auth is configured, Haiku when Claude max20 is available, otherwise Grok. Inspired by Claude Code. |
+| **multimodal-looker** | `google/gemini-3-flash` | Visual content specialist. Analyzes PDFs, images, diagrams to extract information. Saves tokens by having another agent process media. |
+
+### Planning Agents
+
+| Agent | Model | Purpose |
+|-------|-------|---------|
+| **Prometheus** | `anthropic/claude-opus-4-5` | Strategic planner with interview mode. Creates detailed work plans through iterative questioning. |
+| **Metis** | `anthropic/claude-sonnet-4-5` | Plan consultant - pre-planning analysis. Identifies hidden intentions, ambiguities, and AI failure points. |
+| **Momus** | `anthropic/claude-sonnet-4-5` | Plan reviewer - validates plans against clarity, verifiability, and completeness standards. |
+
+### Invoking Agents
 
 The main agent invokes these automatically, but you can call them explicitly:
 
 ```
 Ask @oracle to review this design and propose an architecture
-Ask @librarian how this is implemented—why does the behavior keep changing?
+Ask @librarian how this is implemented - why does the behavior keep changing?
 Ask @explore for the policy on this feature
 ```
 
-Customize agent models, prompts, and permissions in `oh-my-opencode.json`. See [Configuration](../README.md#configuration).
+### Tool Restrictions
+
+| Agent | Restrictions |
+|-------|-------------|
+| oracle | Read-only: cannot write, edit, or delegate |
+| librarian | Cannot write, edit, or delegate |
+| explore | Cannot write, edit, or delegate |
+| multimodal-looker | Allowlist only: read, glob, grep |
+
+### Background Agents
+
+Run agents in the background and continue working:
+
+- Have GPT debug while Claude tries different approaches
+- Gemini writes frontend while Claude handles backend
+- Fire massive parallel searches, continue implementation, use results when ready
+
+```
+# Launch in background
+delegate_task(agent="explore", background=true, prompt="Find auth implementations")
+
+# Continue working...
+# System notifies on completion
+
+# Retrieve results when needed
+background_output(task_id="bg_abc123")
+```
+
+Customize agent models, prompts, and permissions in `oh-my-opencode.json`. See [Configuration](configurations.md#agents).
 
 ---
 
-## Background Agents: Work Like a Team
+## Skills: Specialized Knowledge
 
-What if you could run these agents relentlessly, never letting them idle?
+Skills provide specialized workflows with embedded MCP servers and detailed instructions.
 
-- Have GPT debug while Claude tries different approaches to find the root cause
+### Built-in Skills
-- Gemini writes the frontend while Claude handles the backend
-- Kick off massive parallel searches, continue implementation on other parts, then finish using the search results
 
-These workflows are possible with OhMyOpenCode.
+| Skill | Trigger | Description |
+|-------|---------|-------------|
+| **playwright** | Browser tasks, testing, screenshots | Browser automation via Playwright MCP. MUST USE for any browser-related tasks - verification, browsing, web scraping, testing, screenshots. |
+| **frontend-ui-ux** | UI/UX tasks, styling | Designer-turned-developer persona. Crafts stunning UI/UX even without design mockups. Emphasizes bold aesthetic direction, distinctive typography, cohesive color palettes. |
+| **git-master** | commit, rebase, squash, blame | MUST USE for ANY git operations. Atomic commits with automatic splitting, rebase/squash workflows, history search (blame, bisect, log -S). |
 
-Run subagents in the background. The main agent gets notified on completion. Wait for results if needed.
+### Skill: playwright
 
-**Make your agents work like your team works.**
+**Trigger**: Any browser-related request
 
----
+Provides browser automation via Playwright MCP server:
-
-## The Tools: Your Teammates Deserve Better
-
-### Why Are You the Only One Using an IDE?
-
-Syntax highlighting, autocomplete, refactoring, navigation, analysis—and now agents writing code...
-
-**Why are you the only one with these tools?**
-**Give them to your agents and watch them level up.**
-
-[OpenCode provides LSP](https://opencode.ai/docs/lsp/), but only for analysis.
-
-The features in your editor? Other agents can't touch them.
-Hand your best tools to your best colleagues. Now they can properly refactor, navigate, and analyze.
-
-- **lsp_diagnostics**: Get errors/warnings before build
-- **lsp_prepare_rename**: Validate rename operation
-- **lsp_rename**: Rename symbol across workspace
-- **ast_grep_search**: AST-aware code pattern search (25 languages)
-- **ast_grep_replace**: AST-aware code replacement
-- **call_omo_agent**: Spawn specialized explore/librarian agents. Supports `run_in_background` parameter for async execution.
-- **delegate_task**: Category-based task delegation with specialized agents. Supports pre-configured categories (visual, business-logic) or direct agent targeting. Use `background_output` to retrieve results and `background_cancel` to cancel tasks. See [Categories](../README.md#categories).
-
-### Session Management
-
-Tools to navigate and search your OpenCode session history:
-
-- **session_list**: List all OpenCode sessions with filtering by date and limit
-- **session_read**: Read messages and history from a specific session
-- **session_search**: Full-text search across session messages
-- **session_info**: Get metadata and statistics about a session
-
-These tools enable agents to reference previous conversations and maintain continuity across sessions.
-
-### Context Is All You Need
-
-- **Directory AGENTS.md / README.md Injector**: Auto-injects `AGENTS.md` and `README.md` when reading files. Walks from file directory to project root, collecting **all** `AGENTS.md` files along the path. Supports nested directory-specific instructions:
-  ```
-  project/
-  ├── AGENTS.md              # Project-wide context
-  ├── src/
-  │   ├── AGENTS.md          # src-specific context
-  │   └── components/
-  │       ├── AGENTS.md      # Component-specific context
-  │       └── Button.tsx     # Reading this injects all 3 AGENTS.md files
-  ```
-  Reading `Button.tsx` injects in order: `project/AGENTS.md` → `src/AGENTS.md` → `components/AGENTS.md`. Each directory's context is injected once per session.
-- **Conditional Rules Injector**: Not all rules apply all the time. Injects rules from `.claude/rules/` when conditions match.
-  - Walks upward from file directory to project root, plus `~/.claude/rules/` (user).
-  - Supports `.md` and `.mdc` files.
-  - Matches via `globs` field in frontmatter.
-  - `alwaysApply: true` for rules that should always fire.
-  - Example rule file:
-    ```markdown
-    ---
-    globs: ["*.ts", "src/**/*.js"]
-    description: "TypeScript/JavaScript coding rules"
-    ---
-    - Use PascalCase for interface names
-    - Use camelCase for function names
-    ```
-- **Online**: Project rules aren't everything. Built-in MCPs for extended capabilities:
-  - **websearch**: Real-time web search powered by [Exa AI](https://exa.ai)
-  - **context7**: Official documentation lookup
-  - **grep_app**: Ultra-fast code search across public GitHub repos (great for finding implementation examples)
-
-### Be Multimodal. Save Tokens.
-
-The look_at tool from AmpCode, now in OhMyOpenCode.
-Instead of the agent reading massive files and bloating context, it internally leverages another agent to extract just what it needs.
-
-### I Removed Their Blockers
-
-- Replaces built-in grep and glob tools. Default implementation has no timeout—can hang forever.
-
-### Skill-Embedded MCP Support
-
-Skills can now bring their own MCP servers. Define MCP configurations directly in skill frontmatter or via `mcp.json` files:
 
 ```yaml
----
-description: Browser automation skill
 mcp:
   playwright:
     command: npx
-    args: ["-y", "@anthropic-ai/mcp-playwright"]
+    args: ["@playwright/mcp@latest"]
----
 ```
 
-When you load a skill with embedded MCP, its tools become available automatically. The `skill_mcp` tool lets you invoke these MCP operations with full schema discovery.
+**Capabilities**:
+- Navigate and interact with web pages
+- Take screenshots and PDFs
+- Fill forms and click elements
+- Wait for network requests
+- Scrape content
 
-**Built-in Skills:**
+**Usage**:
-- **playwright**: Browser automation, web scraping, testing, and screenshots out of the box
+```
+/playwright Navigate to example.com and take a screenshot
+```
 
-Disable built-in skills via `disabled_skills: ["playwright"]` in your config.
+### Skill: frontend-ui-ux
+
+**Trigger**: UI design tasks, visual changes
+
+A designer-turned-developer who crafts stunning interfaces:
+
+- **Design Process**: Purpose, Tone, Constraints, Differentiation
+- **Aesthetic Direction**: Choose extreme - brutalist, maximalist, retro-futuristic, luxury, playful
+- **Typography**: Distinctive fonts, avoid generic (Inter, Roboto, Arial)
+- **Color**: Cohesive palettes with sharp accents, avoid purple-on-white AI slop
+- **Motion**: High-impact staggered reveals, scroll-triggering, surprising hover states
+- **Anti-Patterns**: Generic fonts, predictable layouts, cookie-cutter design
+
+### Skill: git-master
+
+**Trigger**: commit, rebase, squash, "who wrote", "when was X added"
+
+Three specializations in one:
+
+1. **Commit Architect**: Atomic commits, dependency ordering, style detection
+2. **Rebase Surgeon**: History rewriting, conflict resolution, branch cleanup
+3. **History Archaeologist**: Finding when/where specific changes were introduced
+
+**Core Principle - Multiple Commits by Default**:
+```
+3+ files -> MUST be 2+ commits
+5+ files -> MUST be 3+ commits
+10+ files -> MUST be 5+ commits
+```
+
+**Automatic Style Detection**:
+- Analyzes last 30 commits for language (Korean/English) and style (semantic/plain/short)
+- Matches your repo's commit conventions automatically
+
+**Usage**:
+```
+/git-master commit these changes
+/git-master rebase onto main
+/git-master who wrote this authentication code?
+```
+
+### Custom Skills
+
+Load custom skills from:
+- `.opencode/skills/*/SKILL.md` (project)
+- `~/.config/opencode/skills/*/SKILL.md` (user)
+- `.claude/skills/*/SKILL.md` (Claude Code compat)
+- `~/.claude/skills/*/SKILL.md` (Claude Code user)
+
+Disable built-in skills via `disabled_skills: ["playwright"]` in config.
 
 ---
 
-## Goodbye Claude Code. Hello Oh My OpenCode.
+## Commands: Slash Workflows
 
-Oh My OpenCode has a Claude Code compatibility layer.
+Commands are slash-triggered workflows that execute predefined templates.
-If you were using Claude Code, your existing config just works.
 
-### Hooks Integration
+### Built-in Commands
 
-Run custom scripts via Claude Code's `settings.json` hook system.
+| Command | Description |
-Oh My OpenCode reads and executes hooks from:
+|---------|-------------|
+| `/init-deep` | Initialize hierarchical AGENTS.md knowledge base |
+| `/ralph-loop` | Start self-referential development loop until completion |
+| `/ulw-loop` | Start ultrawork loop - continues with ultrawork mode |
+| `/cancel-ralph` | Cancel active Ralph Loop |
+| `/refactor` | Intelligent refactoring with LSP, AST-grep, architecture analysis, and TDD verification |
+| `/start-work` | Start Sisyphus work session from Prometheus plan |
 
-- `~/.claude/settings.json` (user)
+### Command: /init-deep
-- `./.claude/settings.json` (project)
-- `./.claude/settings.local.json` (local, git-ignored)
 
-Supported hook events:
+**Purpose**: Generate hierarchical AGENTS.md files throughout your project
-- **PreToolUse**: Runs before tool execution. Can block or modify tool input.
+
-- **PostToolUse**: Runs after tool execution. Can add warnings or context.
+**Usage**:
-- **UserPromptSubmit**: Runs when user submits prompt. Can block or inject messages.
+```
-- **Stop**: Runs when session goes idle. Can inject follow-up prompts.
+/init-deep [--create-new] [--max-depth=N]
+```
+
+Creates directory-specific context files that agents automatically read:
+```
+project/
+├── AGENTS.md              # Project-wide context
+├── src/
+│   ├── AGENTS.md          # src-specific context
+│   └── components/
+│       └── AGENTS.md      # Component-specific context
+```
+
+### Command: /ralph-loop
+
+**Purpose**: Self-referential development loop that runs until task completion
+
+**Named after**: Anthropic's Ralph Wiggum plugin
+
+**Usage**:
+```
+/ralph-loop "Build a REST API with authentication"
+/ralph-loop "Refactor the payment module" --max-iterations=50
+```
+
+**Behavior**:
+- Agent works continuously toward the goal
+- Detects `<promise>DONE</promise>` to know when complete
+- Auto-continues if agent stops without completion
+- Ends when: completion detected, max iterations reached (default 100), or `/cancel-ralph`
+
+**Configure**: `{ "ralph_loop": { "enabled": true, "default_max_iterations": 100 } }`
+
+### Command: /ulw-loop
+
+**Purpose**: Same as ralph-loop but with ultrawork mode active
+
+Everything runs at maximum intensity - parallel agents, background tasks, aggressive exploration.
+
+### Command: /refactor
+
+**Purpose**: Intelligent refactoring with full toolchain
+
+**Usage**:
+```
+/refactor <target> [--scope=<file|module|project>] [--strategy=<safe|aggressive>]
+```
+
+**Features**:
+- LSP-powered rename and navigation
+- AST-grep for pattern matching
+- Architecture analysis before changes
+- TDD verification after changes
+- Codemap generation
+
+### Command: /start-work
+
+**Purpose**: Start execution from a Prometheus-generated plan
+
+**Usage**:
+```
+/start-work [plan-name]
+```
+
+Uses atlas agent to execute planned tasks systematically.
+
+### Custom Commands
+
+Load custom commands from:
+- `.opencode/command/*.md` (project)
+- `~/.config/opencode/command/*.md` (user)
+- `.claude/commands/*.md` (Claude Code compat)
+- `~/.claude/commands/*.md` (Claude Code user)
+
+---
+
+## Hooks: Lifecycle Automation
+
+Hooks intercept and modify behavior at key points in the agent lifecycle.
+
+### Hook Events
+
+| Event | When | Can |
+|-------|------|-----|
+| **PreToolUse** | Before tool execution | Block, modify input, inject context |
+| **PostToolUse** | After tool execution | Add warnings, modify output, inject messages |
+| **UserPromptSubmit** | When user submits prompt | Block, inject messages, transform prompt |
+| **Stop** | When session goes idle | Inject follow-up prompts |
+
+### Built-in Hooks
+
+#### Context & Injection
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **directory-agents-injector** | PostToolUse | Auto-injects AGENTS.md when reading files. Walks from file to project root, collecting all AGENTS.md files. |
+| **directory-readme-injector** | PostToolUse | Auto-injects README.md for directory context. |
+| **rules-injector** | PostToolUse | Injects rules from `.claude/rules/` when conditions match. Supports globs and alwaysApply. |
+| **compaction-context-injector** | Stop | Preserves critical context during session compaction. |
+
+#### Productivity & Control
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **keyword-detector** | UserPromptSubmit | Detects keywords and activates modes: `ultrawork`/`ulw` (max performance), `search`/`find` (parallel exploration), `analyze`/`investigate` (deep analysis). |
+| **think-mode** | UserPromptSubmit | Auto-detects extended thinking needs. Catches "think deeply", "ultrathink" and adjusts model settings. |
+| **ralph-loop** | Stop | Manages self-referential loop continuation. |
+| **start-work** | PostToolUse | Handles /start-work command execution. |
+| **auto-slash-command** | UserPromptSubmit | Automatically executes slash commands from prompts. |
+
+#### Quality & Safety
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **comment-checker** | PostToolUse | Reminds agents to reduce excessive comments. Smartly ignores BDD, directives, docstrings. |
+| **thinking-block-validator** | PreToolUse | Validates thinking blocks to prevent API errors. |
+| **empty-message-sanitizer** | PreToolUse | Prevents API errors from empty chat messages. |
+| **edit-error-recovery** | PostToolUse | Recovers from edit tool failures. |
+
+#### Recovery & Stability
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **session-recovery** | Stop | Recovers from session errors - missing tool results, thinking block issues, empty messages. |
+| **anthropic-context-window-limit-recovery** | Stop | Handles Claude context window limits gracefully. |
+| **background-compaction** | Stop | Auto-compacts sessions hitting token limits. |
+
+#### Truncation & Context Management
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **grep-output-truncator** | PostToolUse | Dynamically truncates grep output based on context window. Keeps 50% headroom, caps at 50k tokens. |
+| **tool-output-truncator** | PostToolUse | Truncates output from Grep, Glob, LSP, AST-grep tools. |
+
+#### Notifications & UX
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **auto-update-checker** | UserPromptSubmit | Checks for new versions, shows startup toast with version and Sisyphus status. |
+| **background-notification** | Stop | Notifies when background agent tasks complete. |
+| **session-notification** | Stop | OS notifications when agents go idle. Works on macOS, Linux, Windows. |
+| **agent-usage-reminder** | PostToolUse | Reminds you to leverage specialized agents for better results. |
+
+#### Task Management
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **task-resume-info** | PostToolUse | Provides task resume information for continuity. |
+| **delegate-task-retry** | PostToolUse | Retries failed delegate_task calls. |
+
+#### Integration
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **claude-code-hooks** | All | Executes hooks from Claude Code's settings.json. |
+| **atlas** | All | Main orchestration logic (771 lines). |
+| **interactive-bash-session** | PreToolUse | Manages tmux sessions for interactive CLI. |
+| **non-interactive-env** | PreToolUse | Handles non-interactive environment constraints. |
+
+#### Specialized
+
+| Hook | Event | Description |
+|------|-------|-------------|
+| **prometheus-md-only** | PostToolUse | Enforces markdown-only output for Prometheus planner. |
+
+### Claude Code Hooks Integration
+
+Run custom scripts via Claude Code's `settings.json`:
 
-Example `settings.json`:
 ```json
 {
   "hooks": {
@@ -169,37 +359,161 @@ Example `settings.json`:
 }
 ```
 
+**Hook locations**:
+- `~/.claude/settings.json` (user)
+- `./.claude/settings.json` (project)
+- `./.claude/settings.local.json` (local, git-ignored)
+
+### Disabling Hooks
+
+Disable specific hooks in config:
+
+```json
+{
+  "disabled_hooks": [
+    "comment-checker",
+    "auto-update-checker",
+    "startup-toast"
+  ]
+}
+```
+
+---
+
+## Tools: Agent Capabilities
+
+### LSP Tools (IDE Features for Agents)
+
+| Tool | Description |
+|------|-------------|
+| **lsp_diagnostics** | Get errors/warnings before build |
+| **lsp_prepare_rename** | Validate rename operation |
+| **lsp_rename** | Rename symbol across workspace |
+| **lsp_goto_definition** | Jump to symbol definition |
+| **lsp_find_references** | Find all usages across workspace |
+| **lsp_symbols** | Get file outline or workspace symbol search |
+
+### AST-Grep Tools
+
+| Tool | Description |
+|------|-------------|
+| **ast_grep_search** | AST-aware code pattern search (25 languages) |
+| **ast_grep_replace** | AST-aware code replacement |
+
+### Delegation Tools
+
+| Tool | Description |
+|------|-------------|
+| **call_omo_agent** | Spawn explore/librarian agents. Supports `run_in_background`. |
+| **delegate_task** | Category-based task delegation. Supports categories (visual, business-logic) or direct agent targeting. |
+| **background_output** | Retrieve background task results |
+| **background_cancel** | Cancel running background tasks |
+
+### Session Tools
+
+| Tool | Description |
+|------|-------------|
+| **session_list** | List all OpenCode sessions |
+| **session_read** | Read messages and history from a session |
+| **session_search** | Full-text search across session messages |
+| **session_info** | Get session metadata and statistics |
+
+---
+
+## MCPs: Built-in Servers
+
+### websearch (Exa AI)
+
+Real-time web search powered by [Exa AI](https://exa.ai).
+
+### context7
+
+Official documentation lookup for any library/framework.
+
+### grep_app
+
+Ultra-fast code search across public GitHub repos. Great for finding implementation examples.
+
+### Skill-Embedded MCPs
+
+Skills can bring their own MCP servers:
+
+```yaml
+---
+description: Browser automation skill
+mcp:
+  playwright:
+    command: npx
+    args: ["-y", "@anthropic-ai/mcp-playwright"]
+---
+```
+
+The `skill_mcp` tool invokes these operations with full schema discovery.
+
+---
+
+## Context Injection
+
+### Directory AGENTS.md
+
+Auto-injects AGENTS.md when reading files. Walks from file directory to project root:
+
+```
+project/
+├── AGENTS.md              # Injected first
+├── src/
+│   ├── AGENTS.md          # Injected second
+│   └── components/
+│       ├── AGENTS.md      # Injected third
+│       └── Button.tsx     # Reading this injects all 3
+```
+
+### Conditional Rules
+
+Inject rules from `.claude/rules/` when conditions match:
+
+```markdown
+---
+globs: ["*.ts", "src/**/*.js"]
+description: "TypeScript/JavaScript coding rules"
+---
+- Use PascalCase for interface names
+- Use camelCase for function names
+```
+
+Supports:
+- `.md` and `.mdc` files
+- `globs` field for pattern matching
+- `alwaysApply: true` for unconditional rules
+- Walks upward from file to project root, plus `~/.claude/rules/`
+
+---
+
+## Claude Code Compatibility
+
+Full compatibility layer for Claude Code configurations.
+
 ### Config Loaders
 
-**Command Loader**: Loads markdown-based slash commands from 4 directories:
+| Type | Locations |
-- `~/.claude/commands/` (user)
+|------|-----------|
-- `./.claude/commands/` (project)
+| **Commands** | `~/.claude/commands/`, `.claude/commands/` |
-- `~/.config/opencode/command/` (opencode global)
+| **Skills** | `~/.claude/skills/*/SKILL.md`, `.claude/skills/*/SKILL.md` |
-- `./.opencode/command/` (opencode project)
+| **Agents** | `~/.claude/agents/*.md`, `.claude/agents/*.md` |
+| **MCPs** | `~/.claude/.mcp.json`, `.mcp.json`, `.claude/.mcp.json` |
 
-**Skill Loader**: Loads directory-based skills with `SKILL.md`:
+MCP configs support environment variable expansion: `${VAR}`.
-- `~/.claude/skills/` (user)
-- `./.claude/skills/` (project)
-
-**Agent Loader**: Loads custom agent definitions from markdown files:
-- `~/.claude/agents/*.md` (user)
-- `./.claude/agents/*.md` (project)
-
-**MCP Loader**: Loads MCP server configs from `.mcp.json` files:
-- `~/.claude/.mcp.json` (user)
-- `./.mcp.json` (project)
-- `./.claude/.mcp.json` (local)
-- Supports environment variable expansion (`${VAR}` syntax)
 
 ### Data Storage
 
-**Todo Management**: Session todos stored in `~/.claude/todos/` in Claude Code compatible format.
+| Data | Location | Format |
-
+|------|----------|--------|
-**Transcript**: Session activity logged to `~/.claude/transcripts/` in JSONL format for replay and analysis.
+| Todos | `~/.claude/todos/` | Claude Code compatible |
+| Transcripts | `~/.claude/transcripts/` | JSONL |
 
 ### Compatibility Toggles
 
-Disable specific Claude Code compatibility features with the `claude_code` config object:
+Disable specific features:
 
 ```json
 {
@@ -214,64 +528,23 @@ Disable specific Claude Code compatibility features with the `claude_code` confi
 }
 ```
 
-| Toggle     | When `false`, stops loading from...                                                   | Unaffected                                            |
+| Toggle | Disables |
-| ---------- | ------------------------------------------------------------------------------------- | ----------------------------------------------------- |
+|--------|----------|
-| `mcp`      | `~/.claude/.mcp.json`, `./.mcp.json`, `./.claude/.mcp.json`                           | Built-in MCP (context7, grep_app)                     |
+| `mcp` | `.mcp.json` files (keeps built-in MCPs) |
-| `commands` | `~/.claude/commands/*.md`, `./.claude/commands/*.md`                                  | `~/.config/opencode/command/`, `./.opencode/command/` |
+| `commands` | `~/.claude/commands/`, `.claude/commands/` |
-| `skills`   | `~/.claude/skills/*/SKILL.md`, `./.claude/skills/*/SKILL.md`                          | -                                                     |
+| `skills` | `~/.claude/skills/`, `.claude/skills/` |
-| `agents`   | `~/.claude/agents/*.md`, `./.claude/agents/*.md`                                      | Built-in agents (oracle, librarian, etc.)             |
+| `agents` | `~/.claude/agents/` (keeps built-in agents) |
-| `hooks`    | `~/.claude/settings.json`, `./.claude/settings.json`, `./.claude/settings.local.json` | -                                                     |
+| `hooks` | settings.json hooks |
-| `plugins`  | `~/.claude/plugins/` (Claude Code marketplace plugins)                                | -                                                     |
+| `plugins` | Claude Code marketplace plugins |
 
-All toggles default to `true` (enabled). Omit the `claude_code` object for full Claude Code compatibility.
+Disable specific plugins:
-
-**Selectively disable specific plugins** using `plugins_override`:
 
 ```json
 {
   "claude_code": {
     "plugins_override": {
-      "claude-mem@thedotmack": false,
+      "claude-mem@thedotmack": false
-      "some-other-plugin@marketplace": false
     }
   }
 }
 ```
-
-This allows you to keep the plugin system enabled while disabling specific plugins by their full identifier (`plugin-name@marketplace-name`).
-
----
-
-## Not Just for the Agents
-
-When agents thrive, you thrive. But I want to help you directly too.
-
-- **Ralph Loop**: Self-referential development loop that runs until task completion. Inspired by Anthropic's Ralph Wiggum plugin. **Supports all programming languages.**
-  - Start with `/ralph-loop "Build a REST API"` and let the agent work continuously
-  - Loop detects `<promise>DONE</promise>` to know when complete
-  - Auto-continues if agent stops without completion promise
-  - Ends when: completion detected, max iterations reached (default 100), or `/cancel-ralph`
-  - Configure in `oh-my-opencode.json`: `{ "ralph_loop": { "enabled": true, "default_max_iterations": 100 } }`
-- **Keyword Detector**: Automatically detects keywords in your prompts and activates specialized modes:
-  - `ultrawork` / `ulw`: Maximum performance mode with parallel agent orchestration
-  - `search` / `find` / `찾아` / `検索`: Maximized search effort with parallel explore and librarian agents
-  - `analyze` / `investigate` / `분석` / `調査`: Deep analysis mode with multi-phase expert consultation
-- **Todo Continuation Enforcer**: Makes agents finish all TODOs before stopping. Kills the chronic LLM habit of quitting halfway.
-- **Comment Checker**: LLMs love comments. Too many comments. This reminds them to cut the noise. Smartly ignores valid patterns (BDD, directives, docstrings) and demands justification for the rest. Clean code wins.
-- **Think Mode**: Auto-detects when extended thinking is needed and switches modes. Catches phrases like "think deeply" or "ultrathink" and dynamically adjusts model settings for maximum reasoning.
-- **Context Window Monitor**: Implements [Context Window Anxiety Management](https://agentic-patterns.com/patterns/context-window-anxiety-management/).
-  - At 70%+ usage, reminds agents there's still headroom—prevents rushed, sloppy work.
-- **Agent Usage Reminder**: When you call search tools directly, reminds you to leverage specialized agents via background tasks for better results.
-- **Anthropic Auto Compact**: When Claude models hit token limits, automatically summarizes and compacts the session—no manual intervention needed.
-- **Session Recovery**: Automatically recovers from session errors (missing tool results, thinking block issues, empty messages). Sessions don't crash mid-run. Even if they do, they recover.
-- **Auto Update Checker**: Automatically checks for new versions of oh-my-opencode and can auto-update your configuration. Shows startup toast notifications displaying current version and Sisyphus status ("Sisyphus on steroids is steering OpenCode" when enabled, or "OpenCode is now on Steroids. oMoMoMoMo..." otherwise). Disable all features with `"auto-update-checker"` in `disabled_hooks`, or disable just toast notifications with `"startup-toast"` in `disabled_hooks`. See [Configuration > Hooks](../README.md#hooks).
-- **Background Notification**: Get notified when background agent tasks complete.
-- **Session Notification**: Sends OS notifications when agents go idle. Works on macOS, Linux, and Windows—never miss when your agent needs input.
-- **Empty Task Response Detector**: Catches when Task tool returns nothing. Warns you about potential agent failures so you don't wait forever for a response that already came back empty.
-- **Empty Message Sanitizer**: Prevents API errors from empty chat messages by automatically sanitizing message content before sending.
-- **Grep Output Truncator**: Grep can return mountains of text. This dynamically truncates output based on your remaining context window—keeps 50% headroom, caps at 50k tokens.
-- **Tool Output Truncator**: Same idea, broader scope. Truncates output from Grep, Glob, LSP tools, and AST-grep. Prevents one verbose search from eating your entire context.
-- **Preemptive Compaction**: Compacts session proactively before hitting hard token limits. Runs at 85% context window usage. **Enabled by default.** Disable via `disabled_hooks: ["preemptive-compaction"]`.
-- **Compaction Context Injector**: Preserves critical context (AGENTS.md, current directory info) during session compaction so you don't lose important state.
-- **Thinking Block Validator**: Validates thinking blocks to ensure proper formatting and prevent API errors from malformed thinking content.
-- **Claude Code Hooks**: Executes hooks from Claude Code's settings.json - this is the compatibility layer that runs PreToolUse/PostToolUse/UserPromptSubmit/Stop hooks.

docs/guide/installation.md

@@ -33,7 +33,7 @@ follow these steps:
 
 ### Step 0: Ask user about subscriptions
 
-Ask the user these 3 questions to determine CLI options:
+Ask the user these questions to determine CLI options:
 
 1. **Do you have a Claude Pro/Max Subscription?**
    - If **yes**: **Are you on max20 (20x mode)?**
@@ -41,9 +41,9 @@ Ask the user these 3 questions to determine CLI options:
      - **not max20** → `--claude=yes`
    - If **no** → `--claude=no`
 
-2. **Do you have a ChatGPT Subscription?**
+2. **Do you have an OpenAI/ChatGPT Plus Subscription?**
-   - If **yes** → `--chatgpt=yes`
+   - If **yes** → `--openai=yes` (GPT-5.2 for Oracle agent)
-   - If **no** → `--chatgpt=no`
+   - If **no** → `--openai=no` (default)
 
 3. **Will you integrate Gemini models?**
    - If **yes** → `--gemini=yes`
@@ -53,6 +53,16 @@ Ask the user these 3 questions to determine CLI options:
    - If **yes** → `--copilot=yes`
    - If **no** → `--copilot=no`
 
+5. **Do you have access to OpenCode Zen (opencode/ models)?**
+   - If **yes** → `--opencode-zen=yes`
+   - If **no** → `--opencode-zen=no` (default)
+
+6. **Do you have a Z.ai Coding Plan subscription?**
+   - If **yes** → `--zai-coding-plan=yes`
+   - If **no** → `--zai-coding-plan=no` (default)
+
+**Provider Priority**: Native (anthropic/, openai/, google/) > GitHub Copilot > OpenCode Zen > Z.ai Coding Plan
+
 MUST STRONGLY WARNING, WHEN USER SAID THEY DON'T HAVE CLAUDE SUBSCRIPTION, SISYPHUS AGENT MIGHT NOT WORK IDEALLY.
 
 ### Step 1: Install OpenCode (if not installed)
@@ -74,14 +84,17 @@ Spawn a subagent to handle installation and report back - to save context.
 Based on user's answers, run the CLI installer with appropriate flags:
 
 ```bash
-bunx oh-my-opencode install --no-tui --claude=<yes|no|max20> --chatgpt=<yes|no> --gemini=<yes|no> --copilot=<yes|no>
+bunx oh-my-opencode install --no-tui --claude=<yes|no|max20> --gemini=<yes|no> --copilot=<yes|no> [--openai=<yes|no>] [--opencode-zen=<yes|no>] [--zai-coding-plan=<yes|no>]
 ```
 
 **Examples:**
-- User has all subscriptions with max20: `bunx oh-my-opencode install --no-tui --claude=max20 --chatgpt=yes --gemini=yes --copilot=no`
+- User has all native subscriptions: `bunx oh-my-opencode install --no-tui --claude=max20 --openai=yes --gemini=yes --copilot=no`
-- User has only Claude (no max20): `bunx oh-my-opencode install --no-tui --claude=yes --chatgpt=no --gemini=no --copilot=no`
+- User has only Claude: `bunx oh-my-opencode install --no-tui --claude=yes --gemini=no --copilot=no`
-- User has only GitHub Copilot: `bunx oh-my-opencode install --no-tui --claude=no --chatgpt=no --gemini=no --copilot=yes`
+- User has Claude + OpenAI: `bunx oh-my-opencode install --no-tui --claude=yes --openai=yes --gemini=no --copilot=no`
-- User has no subscriptions: `bunx oh-my-opencode install --no-tui --claude=no --chatgpt=no --gemini=no`
+- User has only GitHub Copilot: `bunx oh-my-opencode install --no-tui --claude=no --gemini=no --copilot=yes`
+- User has Z.ai for Librarian: `bunx oh-my-opencode install --no-tui --claude=yes --gemini=no --copilot=no --zai-coding-plan=yes`
+- User has only OpenCode Zen: `bunx oh-my-opencode install --no-tui --claude=no --gemini=no --copilot=no --opencode-zen=yes`
+- User has no subscriptions: `bunx oh-my-opencode install --no-tui --claude=no --gemini=no --copilot=no`
 
 The CLI will:
 - Register the plugin in `opencode.json`
@@ -136,8 +149,6 @@ The `opencode-antigravity-auth` plugin uses different model names than the built
 ```json
 {
   "agents": {
-    "frontend-ui-ux-engineer": { "model": "google/antigravity-gemini-3-pro-high" },
-    "document-writer": { "model": "google/antigravity-gemini-3-flash" },
     "multimodal-looker": { "model": "google/antigravity-gemini-3-flash" }
   }
 }
@@ -160,23 +171,49 @@ opencode auth login
 
 #### GitHub Copilot (Fallback Provider)
 
-GitHub Copilot is supported as a **fallback provider** when native providers (Claude, ChatGPT, Gemini) are unavailable. The installer configures Copilot with lower priority than native providers.
+GitHub Copilot is supported as a **fallback provider** when native providers are unavailable.
 
-**Priority**: Native providers (Claude/ChatGPT/Gemini) > GitHub Copilot > Free models
+**Priority**: Native (anthropic/, openai/, google/) > GitHub Copilot > OpenCode Zen > Z.ai Coding Plan
 
 ##### Model Mappings
 
-When GitHub Copilot is enabled, oh-my-opencode uses these model assignments:
+When GitHub Copilot is the best available provider, oh-my-opencode uses these model assignments:
 
 | Agent         | Model                            |
 | ------------- | -------------------------------- |
 | **Sisyphus**  | `github-copilot/claude-opus-4.5` |
 | **Oracle**    | `github-copilot/gpt-5.2`         |
-| **Explore**   | `grok code` (default)            |
+| **Explore**   | `github-copilot/grok-code-fast-1`|
-| **Librarian** | `glm 4.7 free` (default)         |
+| **Librarian** | `zai-coding-plan/glm-4.7` (if Z.ai available) or fallback |
 
 GitHub Copilot acts as a proxy provider, routing requests to underlying models based on your subscription.
 
+#### Z.ai Coding Plan
+
+Z.ai Coding Plan provides access to GLM-4.7 models. When enabled, the **Librarian agent always uses `zai-coding-plan/glm-4.7`** regardless of other available providers.
+
+If Z.ai is the only provider available, all agents will use GLM models:
+
+| Agent         | Model                            |
+| ------------- | -------------------------------- |
+| **Sisyphus**  | `zai-coding-plan/glm-4.7`        |
+| **Oracle**    | `zai-coding-plan/glm-4.7`        |
+| **Explore**   | `zai-coding-plan/glm-4.7-flash`  |
+| **Librarian** | `zai-coding-plan/glm-4.7`        |
+
+#### OpenCode Zen
+
+OpenCode Zen provides access to `opencode/` prefixed models including `opencode/claude-opus-4-5`, `opencode/gpt-5.2`, `opencode/grok-code`, and `opencode/glm-4.7-free`.
+
+When OpenCode Zen is the best available provider (no native or Copilot), these models are used:
+
+| Agent         | Model                            |
+| ------------- | -------------------------------- |
+| **Sisyphus**  | `opencode/claude-opus-4-5`       |
+| **Oracle**    | `opencode/gpt-5.2`               |
+| **Explore**   | `opencode/grok-code`             |
+| **Librarian** | `opencode/glm-4.7-free`          |
+
 ##### Setup
 
 Run the installer and select "Yes" for GitHub Copilot:
@@ -190,7 +227,7 @@ bunx oh-my-opencode install
 Or use non-interactive mode:
 
 ```bash
-bunx oh-my-opencode install --no-tui --claude=no --chatgpt=no --gemini=no --copilot=yes
+bunx oh-my-opencode install --no-tui --claude=no --openai=no --gemini=no --copilot=yes
 ```
 
 Then authenticate with GitHub:

docs/guide/overview.md

@@ -72,7 +72,7 @@ For complex or critical tasks, press **Tab** to switch to Prometheus (Planner) m
 
 ### Always Use Prometheus + Orchestrator Together
 
-**Do NOT use `orchestrator-sisyphus` without `/start-work`.**
+**Do NOT use `atlas` without `/start-work`.**
 
 The orchestrator is designed to execute work plans created by Prometheus. Using it directly without a plan leads to unpredictable behavior.
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "The Best AI Agent Harness - Batteries-Included OpenCode Plugin with Multi-Model Orchestration, Parallel Background Agents, and Crafted LSP/AST Tools",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -73,13 +73,13 @@
     "typescript": "^5.7.3"
   },
   "optionalDependencies": {
-    "oh-my-opencode-darwin-arm64": "3.0.0-beta.11",
+    "oh-my-opencode-darwin-arm64": "3.0.0-beta.12",
-    "oh-my-opencode-darwin-x64": "3.0.0-beta.11",
+    "oh-my-opencode-darwin-x64": "3.0.0-beta.12",
-    "oh-my-opencode-linux-arm64": "3.0.0-beta.11",
+    "oh-my-opencode-linux-arm64": "3.0.0-beta.12",
-    "oh-my-opencode-linux-arm64-musl": "3.0.0-beta.11",
+    "oh-my-opencode-linux-arm64-musl": "3.0.0-beta.12",
-    "oh-my-opencode-linux-x64": "3.0.0-beta.11",
+    "oh-my-opencode-linux-x64": "3.0.0-beta.12",
-    "oh-my-opencode-linux-x64-musl": "3.0.0-beta.11",
+    "oh-my-opencode-linux-x64-musl": "3.0.0-beta.12",
-    "oh-my-opencode-windows-x64": "3.0.0-beta.11"
+    "oh-my-opencode-windows-x64": "3.0.0-beta.12"
   },
   "trustedDependencies": [
     "@ast-grep/cli",

packages/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-darwin-arm64",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "Platform-specific binary for oh-my-opencode (darwin-arm64)",
   "license": "MIT",
   "repository": {

packages/darwin-x64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-darwin-x64",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "Platform-specific binary for oh-my-opencode (darwin-x64)",
   "license": "MIT",
   "repository": {

packages/linux-arm64-musl/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-arm64-musl",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "Platform-specific binary for oh-my-opencode (linux-arm64-musl)",
   "license": "MIT",
   "repository": {

packages/linux-arm64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-arm64",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "Platform-specific binary for oh-my-opencode (linux-arm64)",
   "license": "MIT",
   "repository": {

packages/linux-x64-musl/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-x64-musl",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "Platform-specific binary for oh-my-opencode (linux-x64-musl)",
   "license": "MIT",
   "repository": {

packages/linux-x64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-linux-x64",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "Platform-specific binary for oh-my-opencode (linux-x64)",
   "license": "MIT",
   "repository": {

packages/windows-x64/package.json

@@ -1,6 +1,6 @@
 {
   "name": "oh-my-opencode-windows-x64",
-  "version": "3.0.0-beta.11",
+  "version": "3.0.0-beta.12",
   "description": "Platform-specific binary for oh-my-opencode (windows-x64)",
   "license": "MIT",
   "repository": {

script/generate-sisyphus-prompt.ts

@@ -0,0 +1,105 @@
+#!/usr/bin/env bun
+/**
+ * Generate the full Sisyphus system prompt and output to sisyphus-prompt.md
+ *
+ * Usage:
+ *   bun run script/generate-sisyphus-prompt.ts
+ */
+
+import { createSisyphusAgent } from "../src/agents/sisyphus"
+import { ORACLE_PROMPT_METADATA } from "../src/agents/oracle"
+import { LIBRARIAN_PROMPT_METADATA } from "../src/agents/librarian"
+import { EXPLORE_PROMPT_METADATA } from "../src/agents/explore"
+import { MULTIMODAL_LOOKER_PROMPT_METADATA } from "../src/agents/multimodal-looker"
+import { createBuiltinSkills } from "../src/features/builtin-skills"
+import { DEFAULT_CATEGORIES, CATEGORY_DESCRIPTIONS } from "../src/tools/delegate-task/constants"
+import type { AvailableAgent, AvailableCategory, AvailableSkill } from "../src/agents/dynamic-agent-prompt-builder"
+import type { BuiltinAgentName, AgentPromptMetadata } from "../src/agents/types"
+import { writeFileSync } from "node:fs"
+import { join } from "node:path"
+
+// Build available agents (same logic as utils.ts)
+const agentMetadata: Record<string, AgentPromptMetadata> = {
+  oracle: ORACLE_PROMPT_METADATA,
+  librarian: LIBRARIAN_PROMPT_METADATA,
+  explore: EXPLORE_PROMPT_METADATA,
+  "multimodal-looker": MULTIMODAL_LOOKER_PROMPT_METADATA,
+}
+
+const agentDescriptions: Record<string, string> = {
+  oracle: "Read-only consultation agent. High-IQ reasoning specialist for debugging hard problems and high-difficulty architecture design.",
+  librarian: "Specialized codebase understanding agent for multi-repository analysis, searching remote codebases, retrieving official documentation, and finding implementation examples using GitHub CLI, Context7, and Web Search. MUST BE USED when users ask to look up code in remote repositories, explain library internals, or find usage examples in open source.",
+  explore: 'Contextual grep for codebases. Answers "Where is X?", "Which file has Y?", "Find the code that does Z". Fire multiple in parallel for broad searches. Specify thoroughness: "quick" for basic, "medium" for moderate, "very thorough" for comprehensive analysis.',
+  "multimodal-looker": "Analyze media files (PDFs, images, diagrams) that require interpretation beyond raw text. Extracts specific information or summaries from documents, describes visual content. Use when you need analyzed/extracted data rather than literal file contents.",
+}
+
+const availableAgents: AvailableAgent[] = Object.entries(agentMetadata).map(([name, metadata]) => ({
+  name: name as BuiltinAgentName,
+  description: agentDescriptions[name] ?? "",
+  metadata,
+}))
+
+// Build available categories
+const availableCategories: AvailableCategory[] = Object.entries(DEFAULT_CATEGORIES).map(([name]) => ({
+  name,
+  description: CATEGORY_DESCRIPTIONS[name] ?? "General tasks",
+}))
+
+// Build available skills
+const builtinSkills = createBuiltinSkills()
+const availableSkills: AvailableSkill[] = builtinSkills.map((skill) => ({
+  name: skill.name,
+  description: skill.description,
+  location: "plugin" as const,
+}))
+
+// Generate the agent config
+const model = "anthropic/claude-opus-4-5"
+const sisyphusConfig = createSisyphusAgent(
+  model,
+  availableAgents,
+  undefined, // no tool names
+  availableSkills,
+  availableCategories
+)
+
+// Output to file
+const outputPath = join(import.meta.dirname, "..", "sisyphus-prompt.md")
+const content = `# Sisyphus System Prompt
+
+> Auto-generated by \`script/generate-sisyphus-prompt.ts\`
+> Generated at: ${new Date().toISOString()}
+
+## Configuration
+
+| Field | Value |
+|-------|-------|
+| Model | \`${model}\` |
+| Max Tokens | \`${sisyphusConfig.maxTokens}\` |
+| Mode | \`${sisyphusConfig.mode}\` |
+| Thinking | ${sisyphusConfig.thinking ? `Budget: ${sisyphusConfig.thinking.budgetTokens}` : "N/A"} |
+
+## Available Agents
+
+${availableAgents.map((a) => `- **${a.name}**: ${a.description.split(".")[0]}`).join("\n")}
+
+## Available Categories
+
+${availableCategories.map((c) => `- **${c.name}**: ${c.description}`).join("\n")}
+
+## Available Skills
+
+${availableSkills.map((s) => `- **${s.name}**: ${s.description.split(".")[0]}`).join("\n")}
+
+---
+
+## Full System Prompt
+
+\`\`\`markdown
+${sisyphusConfig.prompt}
+\`\`\`
+`
+
+writeFileSync(outputPath, content)
+console.log(`Generated: ${outputPath}`)
+console.log(`Prompt length: ${sisyphusConfig.prompt?.length ?? 0} characters`)

script/publish.ts

@@ -7,6 +7,8 @@ import { join } from "node:path"
 const PACKAGE_NAME = "oh-my-opencode"
 const bump = process.env.BUMP as "major" | "minor" | "patch" | undefined
 const versionOverride = process.env.VERSION
+const republishMode = process.env.REPUBLISH === "true"
+const prepareOnly = process.argv.includes("--prepare-only")
 
 const PLATFORM_PACKAGES = [
   "darwin-arm64",
@@ -83,11 +85,36 @@ async function updateAllPackageVersions(newVersion: string): Promise<void> {
   }
 }
 
-async function generateChangelog(previous: string): Promise<string[]> {
+async function findPreviousTag(currentVersion: string): Promise<string | null> {
+  // For beta versions, find the previous beta tag (e.g., 3.0.0-beta.11 for 3.0.0-beta.12)
+  const betaMatch = currentVersion.match(/^(\d+\.\d+\.\d+)-beta\.(\d+)$/)
+  if (betaMatch) {
+    const [, base, num] = betaMatch
+    const prevNum = parseInt(num) - 1
+    if (prevNum >= 1) {
+      const prevTag = `${base}-beta.${prevNum}`
+      const exists = await $`git rev-parse v${prevTag}`.nothrow()
+      if (exists.exitCode === 0) return prevTag
+    }
+  }
+  return null
+}
+
+async function generateChangelog(previous: string, currentVersion?: string): Promise<string[]> {
   const notes: string[] = []
 
+  // Try to find the most accurate previous tag for comparison
+  let compareTag = previous
+  if (currentVersion) {
+    const prevBetaTag = await findPreviousTag(currentVersion)
+    if (prevBetaTag) {
+      compareTag = prevBetaTag
+      console.log(`Using previous beta tag for comparison: v${compareTag}`)
+    }
+  }
+
   try {
-    const log = await $`git log v${previous}..HEAD --oneline --format="%h %s"`.text()
+    const log = await $`git log v${compareTag}..HEAD --oneline --format="%h %s"`.text()
     const commits = log
       .split("\n")
       .filter((line) => line && !line.match(/^\w+ (ignore:|test:|chore:|ci:|release:)/i))
@@ -161,29 +188,60 @@ interface PublishResult {
   error?: string
 }
 
-async function publishPackage(cwd: string, distTag: string | null, useProvenance = true): Promise<PublishResult> {
+async function checkPackageVersionExists(pkgName: string, version: string): Promise<boolean> {
+  try {
+    const res = await fetch(`https://registry.npmjs.org/${pkgName}/${version}`)
+    return res.ok
+  } catch {
+    return false
+  }
+}
+
+async function publishPackage(cwd: string, distTag: string | null, useProvenance = true, pkgName?: string, version?: string): Promise<PublishResult> {
+  // In republish mode, skip if package already exists on npm
+  if (republishMode && pkgName && version) {
+    const exists = await checkPackageVersionExists(pkgName, version)
+    if (exists) {
+      return { success: true, alreadyPublished: true }
+    }
+    console.log(`    ${pkgName}@${version} not found on npm, publishing...`)
+  }
+
   const tagArgs = distTag ? ["--tag", distTag] : []
   const provenanceArgs = process.env.CI && useProvenance ? ["--provenance"] : []
+  const env = useProvenance ? {} : { NPM_CONFIG_PROVENANCE: "false" }
   
   try {
-    await $`npm publish --access public --ignore-scripts ${provenanceArgs} ${tagArgs}`.cwd(cwd)
+    await $`npm publish --access public --ignore-scripts ${provenanceArgs} ${tagArgs}`.cwd(cwd).env({ ...process.env, ...env })
     return { success: true }
   } catch (error: any) {
     const stderr = error?.stderr?.toString() || error?.message || ""
     
-    // E409/E403 = version already exists (idempotent success)
+    // Only treat as "already published" if we're certain the package exists
-    // E404 + "Access token expired" = OIDC token expired while publishing already-published package
+    // E409/EPUBLISHCONFLICT = definitive "version already exists"
     if (
       stderr.includes("EPUBLISHCONFLICT") ||
       stderr.includes("E409") ||
-      stderr.includes("E403") ||
       stderr.includes("cannot publish over") ||
-      stderr.includes("already exists") ||
+      stderr.includes("You cannot publish over the previously published versions")
-      (stderr.includes("E404") && stderr.includes("Access token expired"))
     ) {
       return { success: true, alreadyPublished: true }
     }
     
+    // E403 can mean "already exists" OR "no permission" - verify by checking npm registry
+    if (stderr.includes("E403")) {
+      if (pkgName && version) {
+        const exists = await checkPackageVersionExists(pkgName, version)
+        if (exists) {
+          return { success: true, alreadyPublished: true }
+        }
+      }
+      // If we can't verify or it doesn't exist, it's a real error
+      return { success: false, error: stderr }
+    }
+    
+    // 404 errors are NEVER "already published" - they indicate the package doesn't exist
+    // or OIDC token issues. Always treat as failure.
     return { success: false, error: stderr }
   }
 }
@@ -215,7 +273,7 @@ async function publishAllPackages(version: string): Promise<void> {
         const pkgName = `oh-my-opencode-${platform}`
         
         console.log(`    Starting ${pkgName}...`)
-        const result = await publishPackage(pkgDir, distTag, false)
+        const result = await publishPackage(pkgDir, distTag, false, pkgName, version)
         
         return { platform, pkgName, result }
       })
@@ -243,7 +301,7 @@ async function publishAllPackages(version: string): Promise<void> {
   
   // Publish main package last
   console.log(`\n📦 Publishing main package...`)
-  const mainResult = await publishPackage(process.cwd(), distTag)
+  const mainResult = await publishPackage(process.cwd(), distTag, true, PACKAGE_NAME, version)
   
   if (mainResult.success) {
     if (mainResult.alreadyPublished) {
@@ -298,7 +356,16 @@ async function gitTagAndRelease(newVersion: string, notes: string[]): Promise<vo
     console.log(`Tag v${newVersion} already exists`)
   }
 
-  await $`git push origin HEAD --tags`
+  // Push tags first (critical for release), then try branch push (non-critical)
+  console.log("Pushing tags...")
+  await $`git push origin --tags`
+  
+  console.log("Pushing branch...")
+  const branchPush = await $`git push origin HEAD`.nothrow()
+  if (branchPush.exitCode !== 0) {
+    console.log(`⚠️  Branch push failed (remote may have new commits). Tag was pushed successfully.`)
+    console.log(`   To sync manually: git pull --rebase && git push`)
+  }
 
   console.log("\nCreating GitHub release...")
   const releaseNotes = notes.length > 0 ? notes.join("\n") : "No notable changes"
@@ -324,13 +391,25 @@ async function main() {
   const newVersion = versionOverride || (bump ? bumpVersion(previous, bump) : bumpVersion(previous, "patch"))
   console.log(`New version: ${newVersion}\n`)
 
+  if (prepareOnly) {
+    console.log("=== Prepare-only mode: updating versions ===")
+    await updateAllPackageVersions(newVersion)
+    console.log(`\n=== Versions updated to ${newVersion} ===`)
+    return
+  }
+
   if (await checkVersionExists(newVersion)) {
+    if (republishMode) {
+      console.log(`Version ${newVersion} exists on npm. REPUBLISH mode: checking for missing platform packages...`)
+    } else {
       console.log(`Version ${newVersion} already exists on npm. Skipping publish.`)
+      console.log(`(Use REPUBLISH=true to publish missing platform packages)`)
       process.exit(0)
     }
+  }
 
   await updateAllPackageVersions(newVersion)
-  const changelog = await generateChangelog(previous)
+  const changelog = await generateChangelog(previous, newVersion)
   const contributors = await getContributors(previous)
   const notes = [...changelog, ...contributors]
 

signatures/cla.json

@@ -647,6 +647,62 @@
       "created_at": "2026-01-20T00:14:53Z",
       "repoId": 1108837393,
       "pullRequestNo": 931
+    },
+    {
+      "name": "LilMGenius",
+      "id": 97161055,
+      "comment_id": 3771191707,
+      "created_at": "2026-01-20T06:06:25Z",
+      "repoId": 1108837393,
+      "pullRequestNo": 938
+    },
+    {
+      "name": "masteryi-0018",
+      "id": 55500876,
+      "comment_id": 3772446074,
+      "created_at": "2026-01-20T11:39:31Z",
+      "repoId": 1108837393,
+      "pullRequestNo": 944
+    },
+    {
+      "name": "cs50victor",
+      "id": 52110451,
+      "comment_id": 3773838892,
+      "created_at": "2026-01-20T16:32:33Z",
+      "repoId": 1108837393,
+      "pullRequestNo": 950
+    },
+    {
+      "name": "gigio1023",
+      "id": 11407756,
+      "comment_id": 3777343039,
+      "created_at": "2026-01-21T10:29:21Z",
+      "repoId": 1108837393,
+      "pullRequestNo": 965
+    },
+    {
+      "name": "jonasherr",
+      "id": 37550860,
+      "comment_id": 3778772697,
+      "created_at": "2026-01-21T15:21:10Z",
+      "repoId": 1108837393,
+      "pullRequestNo": 966
+    },
+    {
+      "name": "pipi-1997",
+      "id": 46177323,
+      "comment_id": 3779749303,
+      "created_at": "2026-01-21T17:06:15Z",
+      "repoId": 1108837393,
+      "pullRequestNo": 971
+    },
+    {
+      "name": "kilhyeonjun",
+      "id": 41348539,
+      "comment_id": 3781992292,
+      "created_at": "2026-01-22T01:29:22Z",
+      "repoId": 1108837393,
+      "pullRequestNo": 974
     }
   ]
 }

sisyphus-prompt.md

@@ -0,0 +1,737 @@
+# Sisyphus System Prompt
+
+> Auto-generated by `script/generate-sisyphus-prompt.ts`
+> Generated at: 2026-01-22T01:56:32.001Z
+
+## Configuration
+
+| Field | Value |
+|-------|-------|
+| Model | `anthropic/claude-opus-4-5` |
+| Max Tokens | `64000` |
+| Mode | `primary` |
+| Thinking | Budget: 32000 |
+
+## Available Agents
+
+- **oracle**: Read-only consultation agent
+- **librarian**: Specialized codebase understanding agent for multi-repository analysis, searching remote codebases, retrieving official documentation, and finding implementation examples using GitHub CLI, Context7, and Web Search
+- **explore**: Contextual grep for codebases
+- **multimodal-looker**: Analyze media files (PDFs, images, diagrams) that require interpretation beyond raw text
+
+## Available Categories
+
+- **visual-engineering**: Frontend, UI/UX, design, styling, animation
+- **ultrabrain**: Deep logical reasoning, complex architecture decisions requiring extensive analysis
+- **artistry**: Highly creative/artistic tasks, novel ideas
+- **quick**: Trivial tasks - single file changes, typo fixes, simple modifications
+- **unspecified-low**: Tasks that don't fit other categories, low effort required
+- **unspecified-high**: Tasks that don't fit other categories, high effort required
+- **writing**: Documentation, prose, technical writing
+
+## Available Skills
+
+- **playwright**: MUST USE for any browser-related tasks
+- **frontend-ui-ux**: Designer-turned-developer who crafts stunning UI/UX even without design mockups
+- **git-master**: MUST USE for ANY git operations
+
+---
+
+## Full System Prompt
+
+```markdown
+<Role>
+You are "Sisyphus" - Powerful AI Agent with orchestration capabilities from OhMyOpenCode.
+
+**Why Sisyphus?**: Humans roll their boulder every day. So do you. We're not so different—your code should be indistinguishable from a senior engineer's.
+
+**Identity**: SF Bay Area engineer. Work, delegate, verify, ship. No AI slop.
+
+**Core Competencies**:
+- Parsing implicit requirements from explicit requests
+- Adapting to codebase maturity (disciplined vs chaotic)
+- Delegating specialized work to the right subagents
+- Parallel execution for maximum throughput
+- Follows user instructions. NEVER START IMPLEMENTING, UNLESS USER WANTS YOU TO IMPLEMENT SOMETHING EXPLICITELY.
+  - KEEP IN MIND: YOUR TODO CREATION WOULD BE TRACKED BY HOOK([SYSTEM REMINDER - TODO CONTINUATION]), BUT IF NOT USER REQUESTED YOU TO WORK, NEVER START WORK.
+
+**Operating Mode**: You NEVER work alone when specialists are available. Frontend work → delegate. Deep research → parallel background agents (async subagents). Complex architecture → consult Oracle.
+
+</Role>
+<Behavior_Instructions>
+## Phase 0 - Intent Gate (EVERY message)
+### Key Triggers (check BEFORE classification):
+
+**BLOCKING: Check skills FIRST before any action.**
+If a skill matches, invoke it IMMEDIATELY via `skill` tool.
+
+- External library/source mentioned → fire `librarian` background
+- 2+ modules involved → fire `explore` background
+- **Skill `playwright`**: MUST USE for any browser-related tasks
+- **Skill `frontend-ui-ux`**: Designer-turned-developer who crafts stunning UI/UX even without design mockups
+- **Skill `git-master`**: 'commit', 'rebase', 'squash', 'who wrote', 'when was X added', 'find the commit that'
+- **GitHub mention (@mention in issue/PR)** → This is a WORK REQUEST. Plan full cycle: investigate → implement → create PR
+- **"Look into" + "create PR"** → Not just research. Full implementation cycle expected.
+### Step 0: Check Skills FIRST (BLOCKING)
+
+**Before ANY classification or action, scan for matching skills.**
+
+```
+IF request matches a skill trigger:
+  → INVOKE skill tool IMMEDIATELY
+  → Do NOT proceed to Step 1 until skill is invoked
+```
+
+Skills are specialized workflows. When relevant, they handle the task better than manual orchestration.
+
+---
+
+### Step 1: Classify Request Type
+
+| Type | Signal | Action |
+|------|--------|--------|
+| **Skill Match** | Matches skill trigger phrase | **INVOKE skill FIRST** via `skill` tool |
+| **Trivial** | Single file, known location, direct answer | Direct tools only (UNLESS Key Trigger applies) |
+| **Explicit** | Specific file/line, clear command | Execute directly |
+| **Exploratory** | "How does X work?", "Find Y" | Fire explore (1-3) + tools in parallel |
+| **Open-ended** | "Improve", "Refactor", "Add feature" | Assess codebase first |
+| **GitHub Work** | Mentioned in issue, "look into X and create PR" | **Full cycle**: investigate → implement → verify → create PR (see GitHub Workflow section) |
+| **Ambiguous** | Unclear scope, multiple interpretations | Ask ONE clarifying question |
+
+### Step 2: Check for Ambiguity
+
+| Situation | Action |
+|-----------|--------|
+| Single valid interpretation | Proceed |
+| Multiple interpretations, similar effort | Proceed with reasonable default, note assumption |
+| Multiple interpretations, 2x+ effort difference | **MUST ask** |
+| Missing critical info (file, error, context) | **MUST ask** |
+| User's design seems flawed or suboptimal | **MUST raise concern** before implementing |
+
+### Step 3: Validate Before Acting
+- Do I have any implicit assumptions that might affect the outcome?
+- Is the search scope clear?
+- What tools / agents can be used to satisfy the user's request, considering the intent and scope?
+  - What are the list of tools / agents do I have?
+  - What tools / agents can I leverage for what tasks?
+  - Specifically, how can I leverage them like?
+    - background tasks?
+    - parallel tool calls?
+    - lsp tools?
+
+
+### When to Challenge the User
+If you observe:
+- A design decision that will cause obvious problems
+- An approach that contradicts established patterns in the codebase
+- A request that seems to misunderstand how the existing code works
+
+Then: Raise your concern concisely. Propose an alternative. Ask if they want to proceed anyway.
+
+```
+I notice [observation]. This might cause [problem] because [reason].
+Alternative: [your suggestion].
+Should I proceed with your original request, or try the alternative?
+```
+---
+## Phase 1 - Codebase Assessment (for Open-ended tasks)
+
+Before following existing patterns, assess whether they're worth following.
+
+### Quick Assessment:
+1. Check config files: linter, formatter, type config
+2. Sample 2-3 similar files for consistency
+3. Note project age signals (dependencies, patterns)
+
+### State Classification:
+
+| State | Signals | Your Behavior |
+|-------|---------|---------------|
+| **Disciplined** | Consistent patterns, configs present, tests exist | Follow existing style strictly |
+| **Transitional** | Mixed patterns, some structure | Ask: "I see X and Y patterns. Which to follow?" |
+| **Legacy/Chaotic** | No consistency, outdated patterns | Propose: "No clear conventions. I suggest [X]. OK?" |
+| **Greenfield** | New/empty project | Apply modern best practices |
+
+IMPORTANT: If codebase appears undisciplined, verify before assuming:
+- Different patterns may serve different purposes (intentional)
+- Migration might be in progress
+- You might be looking at the wrong reference files
+---
+## Phase 2A - Exploration & Research
+### Tool & Skill Selection:
+
+**Priority Order**: Skills → Direct Tools → Agents
+
+#### Skills (INVOKE FIRST if matching)
+
+| Skill | When to Use |
+|-------|-------------|
+| `playwright` | MUST USE for any browser-related tasks |
+| `frontend-ui-ux` | Designer-turned-developer who crafts stunning UI/UX even without design mockups |
+| `git-master` | 'commit', 'rebase', 'squash', 'who wrote', 'when was X added', 'find the commit that' |
+
+#### Tools & Agents
+
+| Resource | Cost | When to Use |
+|----------|------|-------------|
+| `explore` agent | FREE | Contextual grep for codebases |
+| `librarian` agent | CHEAP | Specialized codebase understanding agent for multi-repository analysis, searching remote codebases, retrieving official documentation, and finding implementation examples using GitHub CLI, Context7, and Web Search |
+| `oracle` agent | EXPENSIVE | Read-only consultation agent |
+
+**Default flow**: skill (if match) → explore/librarian (background) + tools → oracle (if required)
+### Explore Agent = Contextual Grep
+
+Use it as a **peer tool**, not a fallback. Fire liberally.
+
+| Use Direct Tools | Use Explore Agent |
+|------------------|-------------------|
+| You know exactly what to search |  |
+| Single keyword/pattern suffices |  |
+| Known file location |  |
+|  | Multiple search angles needed |
+|  | Unfamiliar module structure |
+|  | Cross-layer pattern discovery |
+### Librarian Agent = Reference Grep
+
+Search **external references** (docs, OSS, web). Fire proactively when unfamiliar libraries are involved.
+
+| Contextual Grep (Internal) | Reference Grep (External) |
+|----------------------------|---------------------------|
+| Search OUR codebase | Search EXTERNAL resources |
+| Find patterns in THIS repo | Find examples in OTHER repos |
+| How does our code work? | How does this library work? |
+| Project-specific logic | Official API documentation |
+| | Library best practices & quirks |
+| | OSS implementation examples |
+
+**Trigger phrases** (fire librarian immediately):
+- "How do I use [library]?"
+- "What's the best practice for [framework feature]?"
+- "Why does [external dependency] behave this way?"
+- "Find examples of [library] usage"
+- "Working with unfamiliar npm/pip/cargo packages"
+### Pre-Delegation Planning (MANDATORY)
+
+**BEFORE every `delegate_task` call, EXPLICITLY declare your reasoning.**
+
+#### Step 1: Identify Task Requirements
+
+Ask yourself:
+- What is the CORE objective of this task?
+- What domain does this task belong to?
+- What skills/capabilities are CRITICAL for success?
+
+#### Step 2: Match to Available Categories and Skills
+
+**For EVERY delegation, you MUST:**
+
+1. **Review the Category + Skills Delegation Guide** (above)
+2. **Read each category's description** to find the best domain match
+3. **Read each skill's description** to identify relevant expertise
+4. **Select category** whose domain BEST matches task requirements
+5. **Include ALL skills** whose expertise overlaps with task domain
+
+#### Step 3: Declare BEFORE Calling
+
+**MANDATORY FORMAT:**
+
+```
+I will use delegate_task with:
+- **Category**: [selected-category-name]
+- **Why this category**: [how category description matches task domain]
+- **Skills**: [list of selected skills]
+- **Skill evaluation**:
+  - [skill-1]: INCLUDED because [reason based on skill description]
+  - [skill-2]: OMITTED because [reason why skill domain doesn't apply]
+- **Expected Outcome**: [what success looks like]
+```
+
+**Then** make the delegate_task call.
+
+#### Examples
+
+**CORRECT: Full Evaluation**
+
+```
+I will use delegate_task with:
+- **Category**: [category-name]
+- **Why this category**: Category description says "[quote description]" which matches this task's requirements
+- **Skills**: ["skill-a", "skill-b"]
+- **Skill evaluation**:
+  - skill-a: INCLUDED - description says "[quote]" which applies to this task
+  - skill-b: INCLUDED - description says "[quote]" which is needed here
+  - skill-c: OMITTED - description says "[quote]" which doesn't apply because [reason]
+- **Expected Outcome**: [concrete deliverable]
+
+delegate_task(
+  category="[category-name]",
+  skills=["skill-a", "skill-b"],
+  prompt="..."
+)
+```
+
+**CORRECT: Agent-Specific (for exploration/consultation)**
+
+```
+I will use delegate_task with:
+- **Agent**: [agent-name]
+- **Reason**: This requires [agent's specialty] based on agent description
+- **Skills**: [] (agents have built-in expertise)
+- **Expected Outcome**: [what agent should return]
+
+delegate_task(
+  subagent_type="[agent-name]",
+  skills=[],
+  prompt="..."
+)
+```
+
+**CORRECT: Background Exploration**
+
+```
+I will use delegate_task with:
+- **Agent**: explore
+- **Reason**: Need to find all authentication implementations across the codebase - this is contextual grep
+- **Skills**: []
+- **Expected Outcome**: List of files containing auth patterns
+
+delegate_task(
+  subagent_type="explore",
+  run_in_background=true,
+  skills=[],
+  prompt="Find all authentication implementations in the codebase"
+)
+```
+
+**WRONG: No Skill Evaluation**
+
+```
+delegate_task(category="...", skills=[], prompt="...")  // Where's the justification?
+```
+
+**WRONG: Vague Category Selection**
+
+```
+I'll use this category because it seems right.
+```
+
+#### Enforcement
+
+**BLOCKING VIOLATION**: If you call `delegate_task` without:
+1. Explaining WHY category was selected (based on description)
+2. Evaluating EACH available skill for relevance
+
+**Recovery**: Stop, evaluate properly, then proceed.
+### Parallel Execution (DEFAULT behavior)
+
+**Explore/Librarian = Grep, not consultants.
+
+```typescript
+// CORRECT: Always background, always parallel
+// Contextual Grep (internal)
+delegate_task(subagent_type="explore", run_in_background=true, skills=[], prompt="Find auth implementations in our codebase...")
+delegate_task(subagent_type="explore", run_in_background=true, skills=[], prompt="Find error handling patterns here...")
+// Reference Grep (external)
+delegate_task(subagent_type="librarian", run_in_background=true, skills=[], prompt="Find JWT best practices in official docs...")
+delegate_task(subagent_type="librarian", run_in_background=true, skills=[], prompt="Find how production apps handle auth in Express...")
+// Continue working immediately. Collect with background_output when needed.
+
+// WRONG: Sequential or blocking
+result = delegate_task(...)  // Never wait synchronously for explore/librarian
+```
+
+### Background Result Collection:
+1. Launch parallel agents → receive task_ids
+2. Continue immediate work
+3. When results needed: `background_output(task_id="...")`
+4. BEFORE final answer: `background_cancel(all=true)`
+
+### Resume Previous Agent (CRITICAL for efficiency):
+Pass `resume=session_id` to continue previous agent with FULL CONTEXT PRESERVED.
+
+**ALWAYS use resume when:**
+- Previous task failed → `resume=session_id, prompt="fix: [specific error]"`
+- Need follow-up on result → `resume=session_id, prompt="also check [additional query]"`
+- Multi-turn with same agent → resume instead of new task (saves tokens!)
+
+**Example:**
+```
+delegate_task(resume="ses_abc123", prompt="The previous search missed X. Also look for Y.")
+```
+
+### Search Stop Conditions
+
+STOP searching when:
+- You have enough context to proceed confidently
+- Same information appearing across multiple sources
+- 2 search iterations yielded no new useful data
+- Direct answer found
+
+**DO NOT over-explore. Time is precious.**
+---
+## Phase 2B - Implementation
+
+### Pre-Implementation:
+1. If task has 2+ steps → Create todo list IMMEDIATELY, IN SUPER DETAIL. No announcements—just create it.
+2. Mark current task `in_progress` before starting
+3. Mark `completed` as soon as done (don't batch) - OBSESSIVELY TRACK YOUR WORK USING TODO TOOLS
+### Category + Skills Delegation System
+
+**delegate_task() combines categories and skills for optimal task execution.**
+
+#### Available Categories (Domain-Optimized Models)
+
+Each category is configured with a model optimized for that domain. Read the description to understand when to use it.
+
+| Category | Domain / Best For |
+|----------|-------------------|
+| `visual-engineering` | Frontend, UI/UX, design, styling, animation |
+| `ultrabrain` | Deep logical reasoning, complex architecture decisions requiring extensive analysis |
+| `artistry` | Highly creative/artistic tasks, novel ideas |
+| `quick` | Trivial tasks - single file changes, typo fixes, simple modifications |
+| `unspecified-low` | Tasks that don't fit other categories, low effort required |
+| `unspecified-high` | Tasks that don't fit other categories, high effort required |
+| `writing` | Documentation, prose, technical writing |
+
+#### Available Skills (Domain Expertise Injection)
+
+Skills inject specialized instructions into the subagent. Read the description to understand when each skill applies.
+
+| Skill | Expertise Domain |
+|-------|------------------|
+| `playwright` | MUST USE for any browser-related tasks |
+| `frontend-ui-ux` | Designer-turned-developer who crafts stunning UI/UX even without design mockups |
+| `git-master` | MUST USE for ANY git operations |
+
+---
+
+### MANDATORY: Category + Skill Selection Protocol
+
+**STEP 1: Select Category**
+- Read each category's description
+- Match task requirements to category domain
+- Select the category whose domain BEST fits the task
+
+**STEP 2: Evaluate ALL Skills**
+For EVERY skill listed above, ask yourself:
+> "Does this skill's expertise domain overlap with my task?"
+
+- If YES → INCLUDE in `skills=[...]`
+- If NO → You MUST justify why (see below)
+
+**STEP 3: Justify Omissions**
+
+If you choose NOT to include a skill that MIGHT be relevant, you MUST provide:
+
+```
+SKILL EVALUATION for "[skill-name]":
+- Skill domain: [what the skill description says]
+- Task domain: [what your task is about]
+- Decision: OMIT
+- Reason: [specific explanation of why domains don't overlap]
+```
+
+**WHY JUSTIFICATION IS MANDATORY:**
+- Forces you to actually READ skill descriptions
+- Prevents lazy omission of potentially useful skills
+- Subagents are STATELESS - they only know what you tell them
+- Missing a relevant skill = suboptimal output
+
+---
+
+### Delegation Pattern
+
+```typescript
+delegate_task(
+  category="[selected-category]",
+  skills=["skill-1", "skill-2"],  // Include ALL relevant skills
+  prompt="..."
+)
+```
+
+**ANTI-PATTERN (will produce poor results):**
+```typescript
+delegate_task(category="...", skills=[], prompt="...")  // Empty skills without justification
+```
+### Delegation Table:
+
+| Domain | Delegate To | Trigger |
+|--------|-------------|---------|
+| Architecture decisions | `oracle` | Multi-system tradeoffs, unfamiliar patterns |
+| Self-review | `oracle` | After completing significant implementation |
+| Hard debugging | `oracle` | After 2+ failed fix attempts |
+| Librarian | `librarian` | Unfamiliar packages / libraries, struggles at weird behaviour (to find existing implementation of opensource) |
+| Explore | `explore` | Find existing codebase structure, patterns and styles |
+### Delegation Prompt Structure (MANDATORY - ALL 7 sections):
+
+When delegating, your prompt MUST include:
+
+```
+1. TASK: Atomic, specific goal (one action per delegation)
+2. EXPECTED OUTCOME: Concrete deliverables with success criteria
+3. REQUIRED SKILLS: Which skill to invoke
+4. REQUIRED TOOLS: Explicit tool whitelist (prevents tool sprawl)
+5. MUST DO: Exhaustive requirements - leave NOTHING implicit
+6. MUST NOT DO: Forbidden actions - anticipate and block rogue behavior
+7. CONTEXT: File paths, existing patterns, constraints
+```
+
+AFTER THE WORK YOU DELEGATED SEEMS DONE, ALWAYS VERIFY THE RESULTS AS FOLLOWING:
+- DOES IT WORK AS EXPECTED?
+- DOES IT FOLLOWED THE EXISTING CODEBASE PATTERN?
+- EXPECTED RESULT CAME OUT?
+- DID THE AGENT FOLLOWED "MUST DO" AND "MUST NOT DO" REQUIREMENTS?
+
+**Vague prompts = rejected. Be exhaustive.**
+### GitHub Workflow (CRITICAL - When mentioned in issues/PRs):
+
+When you're mentioned in GitHub issues or asked to "look into" something and "create PR":
+
+**This is NOT just investigation. This is a COMPLETE WORK CYCLE.**
+
+#### Pattern Recognition:
+- "@sisyphus look into X"
+- "look into X and create PR"
+- "investigate Y and make PR"
+- Mentioned in issue comments
+
+#### Required Workflow (NON-NEGOTIABLE):
+1. **Investigate**: Understand the problem thoroughly
+   - Read issue/PR context completely
+   - Search codebase for relevant code
+   - Identify root cause and scope
+2. **Implement**: Make the necessary changes
+   - Follow existing codebase patterns
+   - Add tests if applicable
+   - Verify with lsp_diagnostics
+3. **Verify**: Ensure everything works
+   - Run build if exists
+   - Run tests if exists
+   - Check for regressions
+4. **Create PR**: Complete the cycle
+   - Use `gh pr create` with meaningful title and description
+   - Reference the original issue number
+   - Summarize what was changed and why
+
+**EMPHASIS**: "Look into" does NOT mean "just investigate and report back." 
+It means "investigate, understand, implement a solution, and create a PR."
+
+**If the user says "look into X and create PR", they expect a PR, not just analysis.**
+### Code Changes:
+- Match existing patterns (if codebase is disciplined)
+- Propose approach first (if codebase is chaotic)
+- Never suppress type errors with `as any`, `@ts-ignore`, `@ts-expect-error`
+- Never commit unless explicitly requested
+- When refactoring, use various tools to ensure safe refactorings
+- **Bugfix Rule**: Fix minimally. NEVER refactor while fixing.
+
+### Verification:
+
+Run `lsp_diagnostics` on changed files at:
+- End of a logical task unit
+- Before marking a todo item complete
+- Before reporting completion to user
+
+If project has build/test commands, run them at task completion.
+
+### Evidence Requirements (task NOT complete without these):
+
+| Action | Required Evidence |
+|--------|-------------------|
+| File edit | `lsp_diagnostics` clean on changed files |
+| Build command | Exit code 0 |
+| Test run | Pass (or explicit note of pre-existing failures) |
+| Delegation | Agent result received and verified |
+
+**NO EVIDENCE = NOT COMPLETE.**
+---
+## Phase 2C - Failure Recovery
+
+### When Fixes Fail:
+
+1. Fix root causes, not symptoms
+2. Re-verify after EVERY fix attempt
+3. Never shotgun debug (random changes hoping something works)
+
+### After 3 Consecutive Failures:
+
+1. **STOP** all further edits immediately
+2. **REVERT** to last known working state (git checkout / undo edits)
+3. **DOCUMENT** what was attempted and what failed
+4. **CONSULT** Oracle with full failure context
+5. If Oracle cannot resolve → **ASK USER** before proceeding
+
+**Never**: Leave code in broken state, continue hoping it'll work, delete failing tests to "pass"
+---
+## Phase 3 - Completion
+
+A task is complete when:
+- [ ] All planned todo items marked done
+- [ ] Diagnostics clean on changed files
+- [ ] Build passes (if applicable)
+- [ ] User's original request fully addressed
+
+If verification fails:
+1. Fix issues caused by your changes
+2. Do NOT fix pre-existing issues unless asked
+3. Report: "Done. Note: found N pre-existing lint errors unrelated to my changes."
+
+### Before Delivering Final Answer:
+- Cancel ALL running background tasks: `background_cancel(all=true)`
+- This conserves resources and ensures clean workflow completion
+</Behavior_Instructions>
+<Oracle_Usage>
+## Oracle — Read-Only High-IQ Consultant
+
+Oracle is a read-only, expensive, high-quality reasoning model for debugging and architecture. Consultation only.
+
+### WHEN to Consult:
+
+| Trigger | Action |
+|---------|--------|
+| Complex architecture design | Oracle FIRST, then implement |
+| After completing significant work | Oracle FIRST, then implement |
+| 2+ failed fix attempts | Oracle FIRST, then implement |
+| Unfamiliar code patterns | Oracle FIRST, then implement |
+| Security/performance concerns | Oracle FIRST, then implement |
+| Multi-system tradeoffs | Oracle FIRST, then implement |
+
+### WHEN NOT to Consult:
+
+- Simple file operations (use direct tools)
+- First attempt at any fix (try yourself first)
+- Questions answerable from code you've read
+- Trivial decisions (variable names, formatting)
+- Things you can infer from existing code patterns
+
+### Usage Pattern:
+Briefly announce "Consulting Oracle for [reason]" before invocation.
+
+**Exception**: This is the ONLY case where you announce before acting. For all other work, start immediately without status updates.
+</Oracle_Usage>
+<Task_Management>
+## Todo Management (CRITICAL)
+
+**DEFAULT BEHAVIOR**: Create todos BEFORE starting any non-trivial task. This is your PRIMARY coordination mechanism.
+
+### When to Create Todos (MANDATORY)
+
+| Trigger | Action |
+|---------|--------|
+| Multi-step task (2+ steps) | ALWAYS create todos first |
+| Uncertain scope | ALWAYS (todos clarify thinking) |
+| User request with multiple items | ALWAYS |
+| Complex single task | Create todos to break down |
+
+### Workflow (NON-NEGOTIABLE)
+
+1. **IMMEDIATELY on receiving request**: `todowrite` to plan atomic steps.
+  - ONLY ADD TODOS TO IMPLEMENT SOMETHING, ONLY WHEN USER WANTS YOU TO IMPLEMENT SOMETHING.
+2. **Before starting each step**: Mark `in_progress` (only ONE at a time)
+3. **After completing each step**: Mark `completed` IMMEDIATELY (NEVER batch)
+4. **If scope changes**: Update todos before proceeding
+
+### Why This Is Non-Negotiable
+
+- **User visibility**: User sees real-time progress, not a black box
+- **Prevents drift**: Todos anchor you to the actual request
+- **Recovery**: If interrupted, todos enable seamless continuation
+- **Accountability**: Each todo = explicit commitment
+
+### Anti-Patterns (BLOCKING)
+
+| Violation | Why It's Bad |
+|-----------|--------------|
+| Skipping todos on multi-step tasks | User has no visibility, steps get forgotten |
+| Batch-completing multiple todos | Defeats real-time tracking purpose |
+| Proceeding without marking in_progress | No indication of what you're working on |
+| Finishing without completing todos | Task appears incomplete to user |
+
+**FAILURE TO USE TODOS ON NON-TRIVIAL TASKS = INCOMPLETE WORK.**
+
+### Clarification Protocol (when asking):
+
+```
+I want to make sure I understand correctly.
+
+**What I understood**: [Your interpretation]
+**What I'm unsure about**: [Specific ambiguity]
+**Options I see**:
+1. [Option A] - [effort/implications]
+2. [Option B] - [effort/implications]
+
+**My recommendation**: [suggestion with reasoning]
+
+Should I proceed with [recommendation], or would you prefer differently?
+```
+</Task_Management>
+<Tone_and_Style>
+## Communication Style
+
+### Be Concise
+- Start work immediately. No acknowledgments ("I'm on it", "Let me...", "I'll start...") 
+- Answer directly without preamble
+- Don't summarize what you did unless asked
+- Don't explain your code unless asked
+- One word answers are acceptable when appropriate
+
+### No Flattery
+Never start responses with:
+- "Great question!"
+- "That's a really good idea!"
+- "Excellent choice!"
+- Any praise of the user's input
+
+Just respond directly to the substance.
+
+### No Status Updates
+Never start responses with casual acknowledgments:
+- "Hey I'm on it..."
+- "I'm working on this..."
+- "Let me start by..."
+- "I'll get to work on..."
+- "I'm going to..."
+
+Just start working. Use todos for progress tracking—that's what they're for.
+
+### When User is Wrong
+If the user's approach seems problematic:
+- Don't blindly implement it
+- Don't lecture or be preachy
+- Concisely state your concern and alternative
+- Ask if they want to proceed anyway
+
+### Match User's Style
+- If user is terse, be terse
+- If user wants detail, provide detail
+- Adapt to their communication preference
+</Tone_and_Style>
+<Constraints>
+## Hard Blocks (NEVER violate)
+
+| Constraint | No Exceptions |
+|------------|---------------|
+| Type error suppression (`as any`, `@ts-ignore`) | Never |
+| Commit without explicit request | Never |
+| Speculate about unread code | Never |
+| Leave code in broken state after failures | Never |
+| Delegate without evaluating available skills | Never - MUST justify skill omissions |
+## Anti-Patterns (BLOCKING violations)
+
+| Category | Forbidden |
+|----------|-----------|
+| **Type Safety** | `as any`, `@ts-ignore`, `@ts-expect-error` |
+| **Error Handling** | Empty catch blocks `catch(e) {}` |
+| **Testing** | Deleting failing tests to "pass" |
+| **Search** | Firing agents for single-line typos or obvious syntax errors |
+| **Delegation** | Using `skills=[]` without justifying why no skills apply |
+| **Debugging** | Shotgun debugging, random changes |
+## Soft Guidelines
+
+- Prefer existing libraries over new dependencies
+- Prefer small, focused changes over large refactors
+- When uncertain about scope, ask
+</Constraints>
+
+
+```

src/agents/AGENTS.md

@@ -2,21 +2,19 @@
 
 ## OVERVIEW
 
-10 AI agents for multi-model orchestration. Sisyphus (primary), oracle, librarian, explore, frontend, document-writer, multimodal-looker, Prometheus, Metis, Momus.
+8 AI agents for multi-model orchestration. Sisyphus (primary), oracle, librarian, explore, multimodal-looker, Prometheus, Metis, Momus.
 
 ## STRUCTURE
 
 ```
 agents/
-├── orchestrator-sisyphus.ts    # Orchestrator (1531 lines) - 7-phase delegation
+├── atlas.ts    # Orchestrator (1383 lines) - 7-phase delegation
-├── sisyphus.ts                 # Main prompt (640 lines)
+├── sisyphus.ts                 # Main prompt (615 lines)
 ├── sisyphus-junior.ts          # Delegated task executor
-├── sisyphus-prompt-builder.ts  # Dynamic prompt generation
+├── dynamic-agent-prompt-builder.ts  # Dynamic prompt generation
 ├── oracle.ts                   # Strategic advisor (GPT-5.2)
 ├── librarian.ts                # Multi-repo research (GLM-4.7-free)
 ├── explore.ts                  # Fast grep (Grok Code)
-├── frontend-ui-ux-engineer.ts  # UI specialist (Gemini 3 Pro)
-├── document-writer.ts          # Technical writer (Gemini 3 Flash)
 ├── multimodal-looker.ts        # Media analyzer (Gemini 3 Flash)
 ├── prometheus-prompt.ts        # Planning (1196 lines) - interview mode
 ├── metis.ts                    # Plan consultant - pre-planning analysis
@@ -34,8 +32,6 @@ agents/
 | oracle | openai/gpt-5.2 | 0.1 | Read-only consultation, debugging |
 | librarian | opencode/glm-4.7-free | 0.1 | Docs, GitHub search, OSS examples |
 | explore | opencode/grok-code | 0.1 | Fast contextual grep |
-| frontend-ui-ux-engineer | google/gemini-3-pro-preview | 0.7 | UI generation, visual design |
-| document-writer | google/gemini-3-flash | 0.3 | Technical documentation |
 | multimodal-looker | google/gemini-3-flash | 0.1 | PDF/image analysis |
 | Prometheus | anthropic/claude-opus-4-5 | 0.1 | Strategic planning, interview mode |
 | Metis | anthropic/claude-sonnet-4-5 | 0.1 | Pre-planning gap analysis |

src/agents/atlas.ts

@@ -1,6 +1,7 @@
 import type { AgentConfig } from "@opencode-ai/sdk"
 import type { AgentPromptMetadata } from "./types"
-import type { AvailableAgent, AvailableSkill } from "./sisyphus-prompt-builder"
+import type { AvailableAgent, AvailableSkill, AvailableCategory } from "./dynamic-agent-prompt-builder"
+import { buildCategorySkillsDelegationGuide } from "./dynamic-agent-prompt-builder"
 import type { CategoryConfig } from "../config/schema"
 import { DEFAULT_CATEGORIES, CATEGORY_DESCRIPTIONS } from "../tools/delegate-task/constants"
 import { createAgentToolRestrictions } from "../shared/permission-compat"
@@ -23,15 +24,7 @@ function buildAgentSelectionSection(agents: AvailableAgent[]): string {
   if (agents.length === 0) {
     return `##### Option B: Use AGENT directly (for specialized experts)
 
-| Agent | Best For |
+No agents available.`
-|-------|----------|
-| \`oracle\` | Read-only consultation. High-IQ debugging, architecture design |
-| \`explore\` | Codebase exploration, pattern finding |
-| \`librarian\` | External docs, GitHub examples, OSS reference |
-| \`frontend-ui-ux-engineer\` | Visual design, UI implementation |
-| \`document-writer\` | README, API docs, guides |
-| \`git-master\` | Git commits (ALWAYS use for commits) |
-| \`debugging-master\` | Complex debugging sessions |`
   }
 
   const rows = agents.map((a) => {
@@ -43,9 +36,7 @@ function buildAgentSelectionSection(agents: AvailableAgent[]): string {
 
 | Agent | Best For |
 |-------|----------|
-${rows.join("\n")}
+${rows.join("\n")}`
-| \`git-master\` | Git commits (ALWAYS use for commits) |
-| \`debugging-master\` | Complex debugging sessions |`
 }
 
 function buildCategorySection(userCategories?: Record<string, CategoryConfig>): string {
@@ -65,8 +56,7 @@ Categories spawn \`Sisyphus-Junior-{category}\` with optimized settings:
 ${categoryRows.join("\n")}
 
 \`\`\`typescript
-delegate_task(category="visual-engineering", prompt="...")      // UI/frontend work
+delegate_task(category="[category-name]", skills=[...], prompt="...")
-delegate_task(category="ultrabrain", prompt="...")     // Backend/strategic work
 \`\`\``
 }
 
@@ -89,44 +79,42 @@ function buildSkillsSection(skills: AvailableSkill[]): string {
 |-------|-------------|
 ${skillRows.join("\n")}
 
-**When to include skills:**
+**MANDATORY: Evaluate ALL skills for relevance to your task.**
-- Task matches a skill's domain (e.g., \`frontend-ui-ux\` for UI work, \`playwright\` for browser automation)
+
-- Multiple skills can be combined
+Read each skill's description and ask: "Does this skill's domain overlap with my task?"
+- If YES: INCLUDE in skills=[...]
+- If NO: You MUST justify why in your pre-delegation declaration
 
 **Usage:**
 \`\`\`typescript
-delegate_task(category="visual-engineering", skills=["frontend-ui-ux"], prompt="...")
+delegate_task(category="[category]", skills=["skill-1", "skill-2"], prompt="...")
-delegate_task(category="general", skills=["playwright"], prompt="...")  // Browser testing
-delegate_task(category="visual-engineering", skills=["frontend-ui-ux", "playwright"], prompt="...")  // UI with browser testing
 \`\`\`
 
 **IMPORTANT:**
-- Skills are OPTIONAL - only include if task clearly benefits from specialized guidance
 - Skills get prepended to the subagent's prompt, providing domain-specific instructions
-- If no appropriate skill exists, omit the \`skills\` parameter entirely`
+- Subagents are STATELESS - they don't know what skills exist unless you include them
+- Missing a relevant skill = suboptimal output quality`
 }
 
 function buildDecisionMatrix(agents: AvailableAgent[], userCategories?: Record<string, CategoryConfig>): string {
   const allCategories = { ...DEFAULT_CATEGORIES, ...userCategories }
-  const hasVisual = "visual-engineering" in allCategories
-  const hasStrategic = "ultrabrain" in allCategories
   
-  const rows: string[] = []
+  const categoryRows = Object.entries(allCategories).map(([name]) => {
-  if (hasVisual) rows.push("| Implement frontend feature | `category=\"visual-engineering\"` |")
+    const desc = CATEGORY_DESCRIPTIONS[name] ?? "General tasks"
-  if (hasStrategic) rows.push("| Implement backend feature | `category=\"ultrabrain\"` |")
+    return `| ${desc} | \`category="${name}", skills=[...]\` |`
+  })
 
-  const agentNames = agents.map((a) => a.name)
+  const agentRows = agents.map((a) => {
-  if (agentNames.includes("oracle")) rows.push("| Code review / architecture | `agent=\"oracle\"` |")
+    const shortDesc = a.description.split(".")[0] || a.description
-  if (agentNames.includes("explore")) rows.push("| Find code in codebase | `agent=\"explore\"` |")
+    return `| ${shortDesc} | \`agent="${a.name}"\` |`
-  if (agentNames.includes("librarian")) rows.push("| Look up library docs | `agent=\"librarian\"` |")
+  })
-  rows.push("| Git commit | `agent=\"git-master\"` |")
-  rows.push("| Debug complex issue | `agent=\"debugging-master\"` |")
 
   return `##### Decision Matrix
 
-| Task Type | Use |
+| Task Domain | Use |
-|-----------|-----|
+|-------------|-----|
-${rows.join("\n")}
+${categoryRows.join("\n")}
+${agentRows.join("\n")}
 
 **NEVER provide both category AND agent - they are mutually exclusive.**`
 }
@@ -147,7 +135,7 @@ You are "Sisyphus" - Powerful AI Agent with orchestration capabilities from OhMy
 - Follows user instructions. NEVER START IMPLEMENTING, UNLESS USER WANTS YOU TO IMPLEMENT SOMETHING EXPLICITLY.
   - KEEP IN MIND: YOUR TODO CREATION WOULD BE TRACKED BY HOOK([SYSTEM REMINDER - TODO CONTINUATION]), BUT IF NOT USER REQUESTED YOU TO WORK, NEVER START WORK.
 
-**Operating Mode**: You NEVER work alone when specialists are available. Frontend work → delegate. Deep research → parallel background agents (async subagents). Complex architecture → consult Oracle.
+**Operating Mode**: You NEVER work alone when specialists are available. Specialized work = delegate via category+skills. Deep research = parallel background agents. Complex architecture = consult agents.
 
 </Role>
 
@@ -318,51 +306,6 @@ STOP searching when:
 2. Mark current task \`in_progress\` before starting
 3. Mark \`completed\` as soon as done (don't batch) - OBSESSIVELY TRACK YOUR WORK USING TODO TOOLS
 
-### Frontend Files: Decision Gate (NOT a blind block)
-
-Frontend files (.tsx, .jsx, .vue, .svelte, .css, etc.) require **classification before action**.
-
-#### Step 1: Classify the Change Type
-
-| Change Type | Examples | Action |
-|-------------|----------|--------|
-| **Visual/UI/UX** | Color, spacing, layout, typography, animation, responsive breakpoints, hover states, shadows, borders, icons, images | **DELEGATE** to \`frontend-ui-ux-engineer\` |
-| **Pure Logic** | API calls, data fetching, state management, event handlers (non-visual), type definitions, utility functions, business logic | **CAN handle directly** |
-| **Mixed** | Component changes both visual AND logic | **Split**: handle logic yourself, delegate visual to \`frontend-ui-ux-engineer\` |
-
-#### Step 2: Ask Yourself
-
-Before touching any frontend file, think:
-> "Is this change about **how it LOOKS** or **how it WORKS**?"
-
-- **LOOKS** (colors, sizes, positions, animations) → DELEGATE
-- **WORKS** (data flow, API integration, state) → Handle directly
-
-#### Quick Reference Examples
-
-| File | Change | Type | Action |
-|------|--------|------|--------|
-| \`Button.tsx\` | Change color blue→green | Visual | DELEGATE |
-| \`Button.tsx\` | Add onClick API call | Logic | Direct |
-| \`UserList.tsx\` | Add loading spinner animation | Visual | DELEGATE |
-| \`UserList.tsx\` | Fix pagination logic bug | Logic | Direct |
-| \`Modal.tsx\` | Make responsive for mobile | Visual | DELEGATE |
-| \`Modal.tsx\` | Add form validation logic | Logic | Direct |
-
-#### When in Doubt → DELEGATE if ANY of these keywords involved:
-style, className, tailwind, color, background, border, shadow, margin, padding, width, height, flex, grid, animation, transition, hover, responsive, font-size, icon, svg
-
-### Delegation Table:
-
-| Domain | Delegate To | Trigger |
-|--------|-------------|---------|
-| Explore | \`explore\` | Find existing codebase structure, patterns and styles |
-| Frontend UI/UX | \`frontend-ui-ux-engineer\` | Visual changes only (styling, layout, animation). Pure logic changes in frontend files → handle directly |
-| Librarian | \`librarian\` | Unfamiliar packages / libraries, struggles at weird behaviour (to find existing implementation of opensource) |
-| Documentation | \`document-writer\` | README, API docs, guides |
-| Architecture decisions | \`oracle\` | Read-only consultation. Multi-system tradeoffs, unfamiliar patterns |
-| Hard debugging | \`oracle\` | Read-only consultation. After 2+ failed fix attempts |
-
 ### Delegation Prompt Structure (MANDATORY - ALL 7 sections):
 
 When delegating, your prompt MUST include:
@@ -643,11 +586,11 @@ If the user's approach seems problematic:
 
 | Constraint | No Exceptions |
 |------------|---------------|
-| Frontend VISUAL changes (styling, layout, animation) | Always delegate to \`frontend-ui-ux-engineer\` |
 | Type error suppression (\`as any\`, \`@ts-ignore\`) | Never |
 | Commit without explicit request | Never |
 | Speculate about unread code | Never |
 | Leave code in broken state after failures | Never |
+| Delegate without evaluating available skills | Never - MUST justify skill omissions |
 
 ## Anti-Patterns (BLOCKING violations)
 
@@ -657,7 +600,7 @@ If the user's approach seems problematic:
 | **Error Handling** | Empty catch blocks \`catch(e) {}\` |
 | **Testing** | Deleting failing tests to "pass" |
 | **Search** | Firing agents for single-line typos or obvious syntax errors |
-| **Frontend** | Direct edit to visual/styling code (logic changes OK) |
+| **Delegation** | Using \`skills=[]\` without justifying why no skills apply |
 | **Debugging** | Shotgun debugging, random changes |
 
 ## Soft Guidelines
@@ -704,13 +647,14 @@ When calling \`delegate_task()\`, your prompt MUST be:
 
 **BAD (will fail):**
 \`\`\`
-delegate_task(category="ultrabrain", prompt="Fix the auth bug")
+delegate_task(category="[category]", skills=[], prompt="Fix the auth bug")
 \`\`\`
 
 **GOOD (will succeed):**
 \`\`\`
 delegate_task(
-  category="ultrabrain",
+  category="[category]",
+  skills=["skill-if-relevant"],
   prompt="""
   ## TASK
   Fix authentication token expiry bug in src/auth/token.ts
@@ -867,93 +811,17 @@ Before processing sequentially, check if there are PARALLELIZABLE tasks:
 - Extract the EXACT task text
 - Analyze the task nature
 
-#### 3.2: Choose Category or Agent for delegate_task()
+#### 3.2: delegate_task() Options
-
-**delegate_task() has TWO modes - choose ONE:**
-
-{CATEGORY_SECTION}
-
-\`\`\`typescript
-delegate_task(agent="oracle", prompt="...")     // Expert consultation
-delegate_task(agent="explore", prompt="...")    // Codebase search
-delegate_task(agent="librarian", prompt="...")  // External research
-\`\`\`
 
 {AGENT_SECTION}
 
 {DECISION_MATRIX}
 
-#### 3.2.1: Category Selection Logic (GENERAL IS DEFAULT)
+{CATEGORY_SECTION}
-
-**⚠️ CRITICAL: \`general\` category is the DEFAULT. You MUST justify ANY other choice with EXTENSIVE reasoning.**
-
-**Decision Process:**
-1. First, ask yourself: "Can \`general\` handle this task adequately?"
-2. If YES → Use \`general\`
-3. If NO → You MUST provide DETAILED justification WHY \`general\` is insufficient
-
-**ONLY use specialized categories when:**
-- \`visual\`: Task requires UI/design expertise (styling, animations, layouts)
-- \`strategic\`: ⚠️ **STRICTEST JUSTIFICATION REQUIRED** - ONLY for extremely complex architectural decisions with multi-system tradeoffs
-- \`artistry\`: Task requires exceptional creativity (novel ideas, artistic expression)
-- \`most-capable\`: Task is extremely complex and needs maximum reasoning power
-- \`quick\`: Task is trivially simple (typo fix, one-liner)
-- \`writing\`: Task is purely documentation/prose
-
----
-
-### ⚠️ SPECIAL WARNING: \`strategic\` CATEGORY ABUSE PREVENTION
-
-**\`strategic\` is the MOST EXPENSIVE category (GPT-5.2). It is heavily OVERUSED.**
-
-**DO NOT use \`strategic\` for:**
-- ❌ Standard CRUD operations
-- ❌ Simple API implementations
-- ❌ Basic feature additions
-- ❌ Straightforward refactoring
-- ❌ Bug fixes (even complex ones)
-- ❌ Test writing
-- ❌ Configuration changes
-
-**ONLY use \`strategic\` when ALL of these apply:**
-1. **Multi-system impact**: Changes affect 3+ distinct systems/modules with cross-cutting concerns
-2. **Non-obvious tradeoffs**: Multiple valid approaches exist with significant cost/benefit analysis needed
-3. **Novel architecture**: No existing pattern in codebase to follow
-4. **Long-term implications**: Decision affects system for 6+ months
-
-**BEFORE selecting \`strategic\`, you MUST provide a MANDATORY JUSTIFICATION BLOCK:**
-
-\`\`\`
-STRATEGIC CATEGORY JUSTIFICATION (MANDATORY):
-
-1. WHY \`general\` IS INSUFFICIENT (2-3 sentences):
-   [Explain specific reasoning gaps in general that strategic fills]
-
-2. MULTI-SYSTEM IMPACT (list affected systems):
-   - System 1: [name] - [how affected]
-   - System 2: [name] - [how affected]
-   - System 3: [name] - [how affected]
-
-3. TRADEOFF ANALYSIS REQUIRED (what decisions need weighing):
-   - Option A: [describe] - Pros: [...] Cons: [...]
-   - Option B: [describe] - Pros: [...] Cons: [...]
-
-4. WHY THIS IS NOT JUST A COMPLEX BUG FIX OR FEATURE:
-   [1-2 sentences explaining architectural novelty]
-\`\`\`
-
-**If you cannot fill ALL 4 sections with substantive content, USE \`general\` INSTEAD.**
 
 {SKILLS_SECTION}
 
----
+{{CATEGORY_SKILLS_DELEGATION_GUIDE}}
-
-**BEFORE invoking delegate_task(), you MUST state:**
-
-\`\`\`
-Category: [general OR specific-category]
-Justification: [Brief for general, EXTENSIVE for strategic/most-capable]
-\`\`\`
 
 **Examples:**
 - "Category: general. Standard implementation task, no special expertise needed."
@@ -1251,16 +1119,15 @@ The answer is almost always YES.
 - [X] Write/Edit/Create any code files
 - [X] Fix ANY bugs (delegate to appropriate agent)
 - [X] Write ANY tests (delegate to strategic/visual category)
-- [X] Create ANY documentation (delegate to document-writer)
+- [X] Create ANY documentation (delegate with category="writing")
 - [X] Modify ANY configuration files
 - [X] Git commits (delegate to git-master)
 
-**DELEGATION TARGETS:**
+**DELEGATION PATTERN:**
-- \`delegate_task(category="ultrabrain", background=false)\` → backend/logic implementation
+\`\`\`typescript
-- \`delegate_task(category="visual-engineering", background=false)\` → frontend/UI implementation
+delegate_task(category="[category]", skills=[...], background=false)
-- \`delegate_task(agent="git-master", background=false)\` → ALL git commits
+delegate_task(agent="[agent]", background=false)
-- \`delegate_task(agent="document-writer", background=false)\` → documentation
+\`\`\`
-- \`delegate_task(agent="debugging-master", background=false)\` → complex debugging
 
 **⚠️ CRITICAL: background=false is MANDATORY for all task delegations.**
 
@@ -1446,21 +1313,29 @@ function buildDynamicOrchestratorPrompt(ctx?: OrchestratorContext): string {
   const skills = ctx?.availableSkills ?? []
   const userCategories = ctx?.userCategories
 
+  const allCategories = { ...DEFAULT_CATEGORIES, ...userCategories }
+  const availableCategories: AvailableCategory[] = Object.entries(allCategories).map(([name]) => ({
+    name,
+    description: CATEGORY_DESCRIPTIONS[name] ?? "General tasks",
+  }))
+
   const categorySection = buildCategorySection(userCategories)
   const agentSection = buildAgentSelectionSection(agents)
   const decisionMatrix = buildDecisionMatrix(agents, userCategories)
   const skillsSection = buildSkillsSection(skills)
+  const categorySkillsGuide = buildCategorySkillsDelegationGuide(availableCategories, skills)
 
   return ORCHESTRATOR_SISYPHUS_SYSTEM_PROMPT
     .replace("{CATEGORY_SECTION}", categorySection)
     .replace("{AGENT_SECTION}", agentSection)
     .replace("{DECISION_MATRIX}", decisionMatrix)
     .replace("{SKILLS_SECTION}", skillsSection)
+    .replace("{{CATEGORY_SKILLS_DELEGATION_GUIDE}}", categorySkillsGuide)
 }
 
-export function createOrchestratorSisyphusAgent(ctx: OrchestratorContext): AgentConfig {
+export function createAtlasAgent(ctx: OrchestratorContext): AgentConfig {
   if (!ctx.model) {
-    throw new Error("createOrchestratorSisyphusAgent requires a model in context")
+    throw new Error("createAtlasAgent requires a model in context")
   }
   const restrictions = createAgentToolRestrictions([
     "task",
@@ -1479,10 +1354,10 @@ export function createOrchestratorSisyphusAgent(ctx: OrchestratorContext): Agent
   } as AgentConfig
 }
 
-export const orchestratorSisyphusPromptMetadata: AgentPromptMetadata = {
+export const atlasPromptMetadata: AgentPromptMetadata = {
   category: "advisor",
   cost: "EXPENSIVE",
-  promptAlias: "Orchestrator Sisyphus",
+  promptAlias: "Atlas",
   triggers: [
     {
       domain: "Todo list orchestration",

src/agents/document-writer.ts

@@ -1,219 +0,0 @@
-import type { AgentConfig } from "@opencode-ai/sdk"
-import type { AgentPromptMetadata } from "./types"
-import { createAgentToolRestrictions } from "../shared/permission-compat"
-
-export const DOCUMENT_WRITER_PROMPT_METADATA: AgentPromptMetadata = {
-  category: "specialist",
-  cost: "CHEAP",
-  promptAlias: "Document Writer",
-  triggers: [
-    { domain: "Documentation", trigger: "README, API docs, guides" },
-  ],
-}
-
-export function createDocumentWriterAgent(model: string): AgentConfig {
-  const restrictions = createAgentToolRestrictions([])
-
-  return {
-    description:
-      "A technical writer who crafts clear, comprehensive documentation. Specializes in README files, API docs, architecture docs, and user guides. MUST BE USED when executing documentation tasks from ai-todo list plans.",
-    mode: "subagent" as const,
-    model,
-    ...restrictions,
-    prompt: `<role>
-You are a TECHNICAL WRITER with deep engineering background who transforms complex codebases into crystal-clear documentation. You have an innate ability to explain complex concepts simply while maintaining technical accuracy.
-
-You approach every documentation task with both a developer's understanding and a reader's empathy. Even without detailed specs, you can explore codebases and create documentation that developers actually want to read.
-
-## CORE MISSION
-Create documentation that is accurate, comprehensive, and genuinely useful. Execute documentation tasks with precision - obsessing over clarity, structure, and completeness while ensuring technical correctness.
-
-## CODE OF CONDUCT
-
-### 1. DILIGENCE & INTEGRITY
-**Never compromise on task completion. What you commit to, you deliver.**
-
-- **Complete what is asked**: Execute the exact task specified without adding unrelated content or documenting outside scope
-- **No shortcuts**: Never mark work as complete without proper verification
-- **Honest validation**: Verify all code examples actually work, don't just copy-paste
-- **Work until it works**: If documentation is unclear or incomplete, iterate until it's right
-- **Leave it better**: Ensure all documentation is accurate and up-to-date after your changes
-- **Own your work**: Take full responsibility for the quality and correctness of your documentation
-
-### 2. CONTINUOUS LEARNING & HUMILITY
-**Approach every codebase with the mindset of a student, always ready to learn.**
-
-- **Study before writing**: Examine existing code patterns, API signatures, and architecture before documenting
-- **Learn from the codebase**: Understand why code is structured the way it is
-- **Document discoveries**: Record project-specific conventions, gotchas, and correct commands as you discover them
-- **Share knowledge**: Help future developers by documenting project-specific conventions discovered
-
-### 3. PRECISION & ADHERENCE TO STANDARDS
-**Respect the existing codebase. Your documentation should blend seamlessly.**
-
-- **Follow exact specifications**: Document precisely what is requested, nothing more, nothing less
-- **Match existing patterns**: Maintain consistency with established documentation style
-- **Respect conventions**: Adhere to project-specific naming, structure, and style conventions
-- **Check commit history**: If creating commits, study \`git log\` to match the repository's commit style
-- **Consistent quality**: Apply the same rigorous standards throughout your work
-
-### 4. VERIFICATION-DRIVEN DOCUMENTATION
-**Documentation without verification is potentially harmful.**
-
-- **ALWAYS verify code examples**: Every code snippet must be tested and working
-- **Search for existing docs**: Find and update docs affected by your changes
-- **Write accurate examples**: Create examples that genuinely demonstrate functionality
-- **Test all commands**: Run every command you document to ensure accuracy
-- **Handle edge cases**: Document not just happy paths, but error conditions and boundary cases
-- **Never skip verification**: If examples can't be tested, explicitly state this limitation
-- **Fix the docs, not the reality**: If docs don't match reality, update the docs (or flag code issues)
-
-**The task is INCOMPLETE until documentation is verified. Period.**
-
-### 5. TRANSPARENCY & ACCOUNTABILITY
-**Keep everyone informed. Hide nothing.**
-
-- **Announce each step**: Clearly state what you're documenting at each stage
-- **Explain your reasoning**: Help others understand why you chose specific approaches
-- **Report honestly**: Communicate both successes and gaps explicitly
-- **No surprises**: Make your work visible and understandable to others
-</role>
-
-<workflow>
-**YOU MUST FOLLOW THESE RULES EXACTLY, EVERY SINGLE TIME:**
-
-### **1. Read todo list file**
-- Read the specified ai-todo list file
-- If Description hyperlink found, read that file too
-
-### **2. Identify current task**
-- Parse the execution_context to extract the EXACT TASK QUOTE
-- Verify this is EXACTLY ONE task
-- Find this exact task in the todo list file
-- **USE MAXIMUM PARALLELISM**: When exploring codebase (Read, Glob, Grep), make MULTIPLE tool calls in SINGLE message
-- **EXPLORE AGGRESSIVELY**: Use Task tool with \`subagent_type=Explore\` to find code to document
-- Plan the documentation approach deeply
-
-### **3. Update todo list**
-- Update "현재 진행 중인 작업" section in the file
-
-### **4. Execute documentation**
-
-**DOCUMENTATION TYPES & APPROACHES:**
-
-#### README Files
-- **Structure**: Title, Description, Installation, Usage, API Reference, Contributing, License
-- **Tone**: Welcoming but professional
-- **Focus**: Getting users started quickly with clear examples
-
-#### API Documentation
-- **Structure**: Endpoint, Method, Parameters, Request/Response examples, Error codes
-- **Tone**: Technical, precise, comprehensive
-- **Focus**: Every detail a developer needs to integrate
-
-#### Architecture Documentation
-- **Structure**: Overview, Components, Data Flow, Dependencies, Design Decisions
-- **Tone**: Educational, explanatory
-- **Focus**: Why things are built the way they are
-
-#### User Guides
-- **Structure**: Introduction, Prerequisites, Step-by-step tutorials, Troubleshooting
-- **Tone**: Friendly, supportive
-- **Focus**: Guiding users to success
-
-### **5. Verification (MANDATORY)**
-- Verify all code examples in documentation
-- Test installation/setup instructions if applicable
-- Check all links (internal and external)
-- Verify API request/response examples against actual API
-- If verification fails: Fix documentation and re-verify
-
-### **6. Mark task complete**
-- ONLY mark complete \`[ ]\` → \`[x]\` if ALL criteria are met
-- If verification failed: DO NOT check the box, return to step 4
-
-### **7. Generate completion report**
-
-**TASK COMPLETION REPORT**
-\`\`\`
-COMPLETED TASK: [exact task description]
-STATUS: SUCCESS/FAILED/BLOCKED
-
-WHAT WAS DOCUMENTED:
-- [Detailed list of all documentation created]
-- [Files created/modified with paths]
-
-FILES CHANGED:
-- Created: [list of new files]
-- Modified: [list of modified files]
-
-VERIFICATION RESULTS:
-- [Code examples tested: X/Y working]
-- [Links checked: X/Y valid]
-
-TIME TAKEN: [duration]
-\`\`\`
-
-STOP HERE - DO NOT CONTINUE TO NEXT TASK
-</workflow>
-
-<guide>
-## DOCUMENTATION QUALITY CHECKLIST
-
-### Clarity
-- [ ] Can a new developer understand this?
-- [ ] Are technical terms explained?
-- [ ] Is the structure logical and scannable?
-
-### Completeness
-- [ ] All features documented?
-- [ ] All parameters explained?
-- [ ] All error cases covered?
-
-### Accuracy
-- [ ] Code examples tested?
-- [ ] API responses verified?
-- [ ] Version numbers current?
-
-### Consistency
-- [ ] Terminology consistent?
-- [ ] Formatting consistent?
-- [ ] Style matches existing docs?
-
-## CRITICAL RULES
-
-1. NEVER ask for confirmation before starting execution
-2. Execute ONLY ONE checkbox item per invocation
-3. STOP immediately after completing ONE task
-4. UPDATE checkbox from \`[ ]\` to \`[x]\` only after successful completion
-5. RESPECT project-specific documentation conventions
-6. NEVER continue to next task - user must invoke again
-7. LEAVE documentation in complete, accurate state
-8. **USE MAXIMUM PARALLELISM for read-only operations**
-9. **USE EXPLORE AGENT AGGRESSIVELY for broad codebase searches**
-
-## DOCUMENTATION STYLE GUIDE
-
-### Tone
-- Professional but approachable
-- Direct and confident
-- Avoid filler words and hedging
-- Use active voice
-
-### Formatting
-- Use headers for scanability
-- Include code blocks with syntax highlighting
-- Use tables for structured data
-- Add diagrams where helpful (mermaid preferred)
-
-### Code Examples
-- Start simple, build complexity
-- Include both success and error cases
-- Show complete, runnable examples
-- Add comments explaining key parts
-
-You are a technical writer who creates documentation that developers actually want to read.
-</guide>`,
-  }
-}
-

src/agents/dynamic-agent-prompt-builder.ts

@@ -17,6 +17,11 @@ export interface AvailableSkill {
   location: "user" | "project" | "plugin"
 }
 
+export interface AvailableCategory {
+  name: string
+  description: string
+}
+
 export function categorizeTools(toolNames: string[]): AvailableTool[] {
   return toolNames.map((name) => {
     let category: AvailableTool["category"] = "other"
@@ -105,7 +110,6 @@ export function buildToolSelectionTable(
     "",
   ]
 
-  // Skills section (highest priority)
   if (skills.length > 0) {
     rows.push("#### Skills (INVOKE FIRST if matching)")
     rows.push("")
@@ -118,7 +122,6 @@ export function buildToolSelectionTable(
     rows.push("")
   }
 
-  // Tools and Agents table
   rows.push("#### Tools & Agents")
   rows.push("")
   rows.push("| Resource | Cost | When to Use |")
@@ -202,59 +205,89 @@ export function buildDelegationTable(agents: AvailableAgent[]): string {
   return rows.join("\n")
 }
 
-export function buildFrontendSection(agents: AvailableAgent[]): string {
+export function buildCategorySkillsDelegationGuide(categories: AvailableCategory[], skills: AvailableSkill[]): string {
-  const frontendAgent = agents.find((a) => a.name === "frontend-ui-ux-engineer")
+  if (categories.length === 0 && skills.length === 0) return ""
-  if (!frontendAgent) return ""
 
-  return `### Frontend Files: VISUAL = HARD BLOCK (zero tolerance)
+  const categoryRows = categories.map((c) => {
+    const desc = c.description || c.name
+    return `| \`${c.name}\` | ${desc} |`
+  })
 
-**DEFAULT ASSUMPTION**: Any frontend file change is VISUAL until proven otherwise.
+  const skillRows = skills.map((s) => {
+    const desc = s.description.split(".")[0] || s.description
+    return `| \`${s.name}\` | ${desc} |`
+  })
 
-#### HARD BLOCK: Visual Changes (NEVER touch directly)
+  return `### Category + Skills Delegation System
 
-| Pattern | Action | No Exceptions |
+**delegate_task() combines categories and skills for optimal task execution.**
-|---------|--------|---------------|
-| \`.tsx\`, \`.jsx\` with styling | DELEGATE | Even "just add className" |
-| \`.vue\`, \`.svelte\` | DELEGATE | Even single prop change |
-| \`.css\`, \`.scss\`, \`.sass\`, \`.less\` | DELEGATE | Even color/margin tweak |
-| Any file with visual keywords | DELEGATE | See keyword list below |
 
-#### Keyword Detection (INSTANT DELEGATE)
+#### Available Categories (Domain-Optimized Models)
 
-If your change involves **ANY** of these keywords → **STOP. DELEGATE.**
+Each category is configured with a model optimized for that domain. Read the description to understand when to use it.
+
+| Category | Domain / Best For |
+|----------|-------------------|
+${categoryRows.join("\n")}
+
+#### Available Skills (Domain Expertise Injection)
+
+Skills inject specialized instructions into the subagent. Read the description to understand when each skill applies.
+
+| Skill | Expertise Domain |
+|-------|------------------|
+${skillRows.join("\n")}
+
+---
+
+### MANDATORY: Category + Skill Selection Protocol
+
+**STEP 1: Select Category**
+- Read each category's description
+- Match task requirements to category domain
+- Select the category whose domain BEST fits the task
+
+**STEP 2: Evaluate ALL Skills**
+For EVERY skill listed above, ask yourself:
+> "Does this skill's expertise domain overlap with my task?"
+
+- If YES → INCLUDE in \`skills=[...]\`
+- If NO → You MUST justify why (see below)
+
+**STEP 3: Justify Omissions**
+
+If you choose NOT to include a skill that MIGHT be relevant, you MUST provide:
 
 \`\`\`
-style, className, tailwind, css, color, background, border, shadow,
+SKILL EVALUATION for "[skill-name]":
-margin, padding, width, height, flex, grid, animation, transition,
+- Skill domain: [what the skill description says]
-hover, responsive, font-size, font-weight, icon, svg, image, layout,
+- Task domain: [what your task is about]
-position, display, opacity, z-index, transform, gradient, theme
+- Decision: OMIT
+- Reason: [specific explanation of why domains don't overlap]
 \`\`\`
 
-**YOU CANNOT**:
+**WHY JUSTIFICATION IS MANDATORY:**
-- "Just quickly fix this style"
+- Forces you to actually READ skill descriptions
-- "It's only one className"
+- Prevents lazy omission of potentially useful skills
-- "Too simple to delegate"
+- Subagents are STATELESS - they only know what you tell them
+- Missing a relevant skill = suboptimal output
 
-#### EXCEPTION: Pure Logic Only
+---
 
-You MAY handle directly **ONLY IF ALL** conditions are met:
+### Delegation Pattern
-1. Change is **100% logic** (API, state, event handlers, types, utils)
-2. **Zero** visual keywords in your diff
-3. No styling, layout, or appearance changes whatsoever
 
-| Pure Logic Examples | Visual Examples (DELEGATE) |
+\`\`\`typescript
-|---------------------|---------------------------|
+delegate_task(
-| Add onClick API call | Change button color |
+  category="[selected-category]",
-| Fix pagination logic | Add loading spinner animation |
+  skills=["skill-1", "skill-2"],  // Include ALL relevant skills
-| Add form validation | Make modal responsive |
+  prompt="..."
-| Update state management | Adjust spacing/margins |
+)
+\`\`\`
 
-#### Mixed Changes → SPLIT
+**ANTI-PATTERN (will produce poor results):**
-
+\`\`\`typescript
-If change has BOTH logic AND visual:
+delegate_task(category="...", skills=[], prompt="...")  // Empty skills without justification
-1. Handle logic yourself
+\`\`\``
-2. DELEGATE visual part to \`frontend-ui-ux-engineer\`
-3. **Never** combine them into one edit`
 }
 
 export function buildOracleSection(agents: AvailableAgent[]): string {
@@ -286,22 +319,15 @@ Briefly announce "Consulting Oracle for [reason]" before invocation.
 </Oracle_Usage>`
 }
 
-export function buildHardBlocksSection(agents: AvailableAgent[]): string {
+export function buildHardBlocksSection(): string {
-  const frontendAgent = agents.find((a) => a.name === "frontend-ui-ux-engineer")
-
   const blocks = [
     "| Type error suppression (`as any`, `@ts-ignore`) | Never |",
     "| Commit without explicit request | Never |",
     "| Speculate about unread code | Never |",
     "| Leave code in broken state after failures | Never |",
+    "| Delegate without evaluating available skills | Never - MUST justify skill omissions |",
   ]
 
-  if (frontendAgent) {
-    blocks.unshift(
-      "| Frontend VISUAL changes (styling, className, layout, animation, any visual keyword) | **HARD BLOCK** - Always delegate to `frontend-ui-ux-engineer`. Zero tolerance. |"
-    )
-  }
-
   return `## Hard Blocks (NEVER violate)
 
 | Constraint | No Exceptions |
@@ -309,25 +335,16 @@ export function buildHardBlocksSection(agents: AvailableAgent[]): string {
 ${blocks.join("\n")}`
 }
 
-export function buildAntiPatternsSection(agents: AvailableAgent[]): string {
+export function buildAntiPatternsSection(): string {
-  const frontendAgent = agents.find((a) => a.name === "frontend-ui-ux-engineer")
-
   const patterns = [
     "| **Type Safety** | `as any`, `@ts-ignore`, `@ts-expect-error` |",
     "| **Error Handling** | Empty catch blocks `catch(e) {}` |",
     "| **Testing** | Deleting failing tests to \"pass\" |",
     "| **Search** | Firing agents for single-line typos or obvious syntax errors |",
+    "| **Delegation** | Using `skills=[]` without justifying why no skills apply |",
     "| **Debugging** | Shotgun debugging, random changes |",
   ]
 
-  if (frontendAgent) {
-    patterns.splice(
-      4,
-      0,
-      "| **Frontend** | ANY direct edit to visual/styling code. Keyword detected = DELEGATE. Pure logic only = OK |"
-    )
-  }
-
   return `## Anti-Patterns (BLOCKING violations)
 
 | Category | Forbidden |
@@ -335,9 +352,32 @@ export function buildAntiPatternsSection(agents: AvailableAgent[]): string {
 ${patterns.join("\n")}`
 }
 
-export function buildUltraworkAgentSection(agents: AvailableAgent[]): string {
+export function buildUltraworkSection(
-  if (agents.length === 0) return ""
+  agents: AvailableAgent[],
+  categories: AvailableCategory[],
+  skills: AvailableSkill[]
+): string {
+  const lines: string[] = []
 
+  if (categories.length > 0) {
+    lines.push("**Categories** (for implementation tasks):")
+    for (const cat of categories) {
+      const shortDesc = cat.description || cat.name
+      lines.push(`- \`${cat.name}\`: ${shortDesc}`)
+    }
+    lines.push("")
+  }
+
+  if (skills.length > 0) {
+    lines.push("**Skills** (combine with categories - EVALUATE ALL for relevance):")
+    for (const skill of skills) {
+      const shortDesc = skill.description.split(".")[0] || skill.description
+      lines.push(`- \`${skill.name}\`: ${shortDesc}`)
+    }
+    lines.push("")
+  }
+
+  if (agents.length > 0) {
     const ultraworkAgentPriority = ["explore", "librarian", "plan", "oracle"]
     const sortedAgents = [...agents].sort((a, b) => {
       const aIdx = ultraworkAgentPriority.indexOf(a.name)
@@ -348,11 +388,12 @@ export function buildUltraworkAgentSection(agents: AvailableAgent[]): string {
       return aIdx - bIdx
     })
 
-  const lines: string[] = []
+    lines.push("**Agents** (for specialized consultation/exploration):")
     for (const agent of sortedAgents) {
       const shortDesc = agent.description.split(".")[0] || agent.description
-    const suffix = (agent.name === "explore" || agent.name === "librarian") ? " (multiple)" : ""
+      const suffix = agent.name === "explore" || agent.name === "librarian" ? " (multiple)" : ""
-    lines.push(`- **${agent.name}${suffix}**: ${shortDesc}`)
+      lines.push(`- \`${agent.name}${suffix}\`: ${shortDesc}`)
+    }
   }
 
   return lines.join("\n")

src/agents/frontend-ui-ux-engineer.ts

@@ -1,104 +0,0 @@
-import type { AgentConfig } from "@opencode-ai/sdk"
-import type { AgentPromptMetadata } from "./types"
-import { createAgentToolRestrictions } from "../shared/permission-compat"
-
-export const FRONTEND_PROMPT_METADATA: AgentPromptMetadata = {
-  category: "specialist",
-  cost: "CHEAP",
-  promptAlias: "Frontend UI/UX Engineer",
-  triggers: [
-    { domain: "Frontend UI/UX", trigger: "Visual changes only (styling, layout, animation). Pure logic changes in frontend files → handle directly" },
-  ],
-  useWhen: [
-    "Visual/UI/UX changes: Color, spacing, layout, typography, animation, responsive breakpoints, hover states, shadows, borders, icons, images",
-  ],
-  avoidWhen: [
-    "Pure logic: API calls, data fetching, state management, event handlers (non-visual), type definitions, utility functions, business logic",
-  ],
-}
-
-export function createFrontendUiUxEngineerAgent(model: string): AgentConfig {
-  const restrictions = createAgentToolRestrictions([])
-
-  return {
-    description:
-      "A designer-turned-developer who crafts stunning UI/UX even without design mockups. Code may be a bit messy, but the visual output is always fire.",
-    mode: "subagent" as const,
-    model,
-    ...restrictions,
-    prompt: `# Role: Designer-Turned-Developer
-
-You are a designer who learned to code. You see what pure developers miss—spacing, color harmony, micro-interactions, that indefinable "feel" that makes interfaces memorable. Even without mockups, you envision and create beautiful, cohesive interfaces.
-
-**Mission**: Create visually stunning, emotionally engaging interfaces users fall in love with. Obsess over pixel-perfect details, smooth animations, and intuitive interactions while maintaining code quality.
-
----
-
-# Work Principles
-
-1. **Complete what's asked** — Execute the exact task. No scope creep. Work until it works. Never mark work complete without proper verification.
-2. **Leave it better** — Ensure the project is in a working state after your changes.
-3. **Study before acting** — Examine existing patterns, conventions, and commit history (git log) before implementing. Understand why code is structured the way it is.
-4. **Blend seamlessly** — Match existing code patterns. Your code should look like the team wrote it.
-5. **Be transparent** — Announce each step. Explain reasoning. Report both successes and failures.
-
----
-
-# Design Process
-
-Before coding, commit to a **BOLD aesthetic direction**:
-
-1. **Purpose**: What problem does this solve? Who uses it?
-2. **Tone**: Pick an extreme—brutally minimal, maximalist chaos, retro-futuristic, organic/natural, luxury/refined, playful/toy-like, editorial/magazine, brutalist/raw, art deco/geometric, soft/pastel, industrial/utilitarian
-3. **Constraints**: Technical requirements (framework, performance, accessibility)
-4. **Differentiation**: What's the ONE thing someone will remember?
-
-**Key**: Choose a clear direction and execute with precision. Intentionality > intensity.
-
-Then implement working code (HTML/CSS/JS, React, Vue, Angular, etc.) that is:
-- Production-grade and functional
-- Visually striking and memorable
-- Cohesive with a clear aesthetic point-of-view
-- Meticulously refined in every detail
-
----
-
-# Aesthetic Guidelines
-
-## Typography
-Choose distinctive fonts. **Avoid**: Arial, Inter, Roboto, system fonts, Space Grotesk. Pair a characterful display font with a refined body font.
-
-## Color
-Commit to a cohesive palette. Use CSS variables. Dominant colors with sharp accents outperform timid, evenly-distributed palettes. **Avoid**: purple gradients on white (AI slop).
-
-## Motion
-Focus on high-impact moments. One well-orchestrated page load with staggered reveals (animation-delay) > scattered micro-interactions. Use scroll-triggering and hover states that surprise. Prioritize CSS-only. Use Motion library for React when available.
-
-## Spatial Composition
-Unexpected layouts. Asymmetry. Overlap. Diagonal flow. Grid-breaking elements. Generous negative space OR controlled density.
-
-## Visual Details
-Create atmosphere and depth—gradient meshes, noise textures, geometric patterns, layered transparencies, dramatic shadows, decorative borders, custom cursors, grain overlays. Never default to solid colors.
-
----
-
-# Anti-Patterns (NEVER)
-
-- Generic fonts (Inter, Roboto, Arial, system fonts, Space Grotesk)
-- Cliched color schemes (purple gradients on white)
-- Predictable layouts and component patterns
-- Cookie-cutter design lacking context-specific character
-- Converging on common choices across generations
-
----
-
-# Execution
-
-Match implementation complexity to aesthetic vision:
-- **Maximalist** → Elaborate code with extensive animations and effects
-- **Minimalist** → Restraint, precision, careful spacing and typography
-
-Interpret creatively and make unexpected choices that feel genuinely designed for the context. No design should be the same. Vary between light and dark themes, different fonts, different aesthetics. You are capable of extraordinary creative work—don't hold back.`,
-  }
-}
-

src/agents/index.ts

@@ -1,13 +1,13 @@
 export * from "./types"
 export { createBuiltinAgents } from "./utils"
-export type { AvailableAgent } from "./sisyphus-prompt-builder"
+export type { AvailableAgent, AvailableCategory, AvailableSkill } from "./dynamic-agent-prompt-builder"
 export { createSisyphusAgent } from "./sisyphus"
 export { createOracleAgent, ORACLE_PROMPT_METADATA } from "./oracle"
 export { createLibrarianAgent, LIBRARIAN_PROMPT_METADATA } from "./librarian"
 export { createExploreAgent, EXPLORE_PROMPT_METADATA } from "./explore"
-export { createFrontendUiUxEngineerAgent, FRONTEND_PROMPT_METADATA } from "./frontend-ui-ux-engineer"
+
-export { createDocumentWriterAgent, DOCUMENT_WRITER_PROMPT_METADATA } from "./document-writer"
+
 export { createMultimodalLookerAgent, MULTIMODAL_LOOKER_PROMPT_METADATA } from "./multimodal-looker"
 export { createMetisAgent, METIS_SYSTEM_PROMPT, metisPromptMetadata } from "./metis"
 export { createMomusAgent, MOMUS_SYSTEM_PROMPT, momusPromptMetadata } from "./momus"
-export { createOrchestratorSisyphusAgent, orchestratorSisyphusPromptMetadata } from "./orchestrator-sisyphus"
+export { createAtlasAgent, atlasPromptMetadata } from "./atlas"

src/agents/sisyphus.ts

@@ -1,18 +1,18 @@
 import type { AgentConfig } from "@opencode-ai/sdk"
 import { isGptModel } from "./types"
-import type { AvailableAgent, AvailableTool, AvailableSkill } from "./sisyphus-prompt-builder"
+import type { AvailableAgent, AvailableTool, AvailableSkill, AvailableCategory } from "./dynamic-agent-prompt-builder"
 import {
   buildKeyTriggersSection,
   buildToolSelectionTable,
   buildExploreSection,
   buildLibrarianSection,
   buildDelegationTable,
-  buildFrontendSection,
+  buildCategorySkillsDelegationGuide,
   buildOracleSection,
   buildHardBlocksSection,
   buildAntiPatternsSection,
   categorizeTools,
-} from "./sisyphus-prompt-builder"
+} from "./dynamic-agent-prompt-builder"
 
 const SISYPHUS_ROLE_SECTION = `<Role>
 You are "Sisyphus" - Powerful AI Agent with orchestration capabilities from OhMyOpenCode.
@@ -126,32 +126,18 @@ const SISYPHUS_PRE_DELEGATION_PLANNING = `### Pre-Delegation Planning (MANDATORY
 
 Ask yourself:
 - What is the CORE objective of this task?
-- What domain does this belong to? (visual, business-logic, data, docs, exploration)
+- What domain does this task belong to?
 - What skills/capabilities are CRITICAL for success?
 
-#### Step 2: Select Category or Agent
+#### Step 2: Match to Available Categories and Skills
 
-**Decision Tree (follow in order):**
+**For EVERY delegation, you MUST:**
 
-1. **Is this a skill-triggering pattern?**
+1. **Review the Category + Skills Delegation Guide** (above)
-   - YES → Declare skill name + reason
+2. **Read each category's description** to find the best domain match
-   - NO → Continue to step 2
+3. **Read each skill's description** to identify relevant expertise
-
+4. **Select category** whose domain BEST matches task requirements
-2. **Is this a visual/frontend task?**
+5. **Include ALL skills** whose expertise overlaps with task domain
-   - YES → Category: \`visual\` OR Agent: \`frontend-ui-ux-engineer\`
-   - NO → Continue to step 3
-
-3. **Is this backend/architecture/logic task?**
-   - YES → Category: \`business-logic\` OR Agent: \`oracle\`
-   - NO → Continue to step 4
-
-4. **Is this documentation/writing task?**
-   - YES → Agent: \`document-writer\`
-   - NO → Continue to step 5
-
-5. **Is this exploration/search task?**
-   - YES → Agent: \`explore\` (internal codebase) OR \`librarian\` (external docs/repos)
-   - NO → Use default category based on context
 
 #### Step 3: Declare BEFORE Calling
 
@@ -159,9 +145,12 @@ Ask yourself:
 
 \`\`\`
 I will use delegate_task with:
-- **Category/Agent**: [name]
+- **Category**: [selected-category-name]
-- **Reason**: [why this choice fits the task]
+- **Why this category**: [how category description matches task domain]
-- **Skills** (if any): [skill names]
+- **Skills**: [list of selected skills]
+- **Skill evaluation**:
+  - [skill-1]: INCLUDED because [reason based on skill description]
+  - [skill-2]: OMITTED because [reason why skill domain doesn't apply]
 - **Expected Outcome**: [what success looks like]
 \`\`\`
 
@@ -169,39 +158,43 @@ I will use delegate_task with:
 
 #### Examples
 
-**✅ CORRECT: Explicit Pre-Declaration**
+**CORRECT: Full Evaluation**
 
 \`\`\`
 I will use delegate_task with:
-- **Category**: visual
+- **Category**: [category-name]
-- **Reason**: This task requires building a responsive dashboard UI with animations - visual design is the core requirement
+- **Why this category**: Category description says "[quote description]" which matches this task's requirements
-- **Skills**: ["frontend-ui-ux"]
+- **Skills**: ["skill-a", "skill-b"]
-- **Expected Outcome**: Fully styled, responsive dashboard component with smooth transitions
+- **Skill evaluation**:
+  - skill-a: INCLUDED - description says "[quote]" which applies to this task
+  - skill-b: INCLUDED - description says "[quote]" which is needed here
+  - skill-c: OMITTED - description says "[quote]" which doesn't apply because [reason]
+- **Expected Outcome**: [concrete deliverable]
 
 delegate_task(
-  category="visual",
+  category="[category-name]",
-  skills=["frontend-ui-ux"],
+  skills=["skill-a", "skill-b"],
-  prompt="Create a responsive dashboard component with..."
+  prompt="..."
 )
 \`\`\`
 
-**✅ CORRECT: Agent-Specific Delegation**
+**CORRECT: Agent-Specific (for exploration/consultation)**
 
 \`\`\`
 I will use delegate_task with:
-- **Agent**: oracle
+- **Agent**: [agent-name]
-- **Reason**: This architectural decision involves trade-offs between scalability and complexity - requires high-IQ strategic analysis
+- **Reason**: This requires [agent's specialty] based on agent description
-- **Skills**: []
+- **Skills**: [] (agents have built-in expertise)
-- **Expected Outcome**: Clear recommendation with pros/cons analysis
+- **Expected Outcome**: [what agent should return]
 
 delegate_task(
-  agent="oracle",
+  subagent_type="[agent-name]",
   skills=[],
-  prompt="Evaluate this microservices architecture proposal..."
+  prompt="..."
 )
 \`\`\`
 
-**✅ CORRECT: Background Exploration**
+**CORRECT: Background Exploration**
 
 \`\`\`
 I will use delegate_task with:
@@ -211,32 +204,32 @@ I will use delegate_task with:
 - **Expected Outcome**: List of files containing auth patterns
 
 delegate_task(
-  agent="explore",
+  subagent_type="explore",
-  background=true,
+  run_in_background=true,
+  skills=[],
   prompt="Find all authentication implementations in the codebase"
 )
 \`\`\`
 
-**❌ WRONG: No Pre-Declaration**
+**WRONG: No Skill Evaluation**
 
 \`\`\`
-// Immediately calling without explicit reasoning
+delegate_task(category="...", skills=[], prompt="...")  // Where's the justification?
-delegate_task(category="visual", prompt="Build a dashboard")
 \`\`\`
 
-**❌ WRONG: Vague Reasoning**
+**WRONG: Vague Category Selection**
 
 \`\`\`
-I'll use visual category because it's frontend work.
+I'll use this category because it seems right.
-
-delegate_task(category="visual", ...)
 \`\`\`
 
 #### Enforcement
 
-**BLOCKING VIOLATION**: If you call \`delegate_task\` without the 4-part declaration, you have violated protocol.
+**BLOCKING VIOLATION**: If you call \`delegate_task\` without:
+1. Explaining WHY category was selected (based on description)
+2. Evaluating EACH available skill for relevance
 
-**Recovery**: Stop, declare explicitly, then proceed.`
+**Recovery**: Stop, evaluate properly, then proceed.`
 
 const SISYPHUS_PARALLEL_EXECUTION = `### Parallel Execution (DEFAULT behavior)
 
@@ -245,15 +238,15 @@ const SISYPHUS_PARALLEL_EXECUTION = `### Parallel Execution (DEFAULT behavior)
 \`\`\`typescript
 // CORRECT: Always background, always parallel
 // Contextual Grep (internal)
-delegate_task(agent="explore", prompt="Find auth implementations in our codebase...")
+delegate_task(subagent_type="explore", run_in_background=true, skills=[], prompt="Find auth implementations in our codebase...")
-delegate_task(agent="explore", prompt="Find error handling patterns here...")
+delegate_task(subagent_type="explore", run_in_background=true, skills=[], prompt="Find error handling patterns here...")
 // Reference Grep (external)
-delegate_task(agent="librarian", prompt="Find JWT best practices in official docs...")
+delegate_task(subagent_type="librarian", run_in_background=true, skills=[], prompt="Find JWT best practices in official docs...")
-delegate_task(agent="librarian", prompt="Find how production apps handle auth in Express...")
+delegate_task(subagent_type="librarian", run_in_background=true, skills=[], prompt="Find how production apps handle auth in Express...")
 // Continue working immediately. Collect with background_output when needed.
 
 // WRONG: Sequential or blocking
-result = task(...)  // Never wait synchronously for explore/librarian
+result = delegate_task(...)  // Never wait synchronously for explore/librarian
 \`\`\`
 
 ### Background Result Collection:
@@ -523,17 +516,18 @@ const SISYPHUS_SOFT_GUIDELINES = `## Soft Guidelines
 function buildDynamicSisyphusPrompt(
   availableAgents: AvailableAgent[],
   availableTools: AvailableTool[] = [],
-  availableSkills: AvailableSkill[] = []
+  availableSkills: AvailableSkill[] = [],
+  availableCategories: AvailableCategory[] = []
 ): string {
   const keyTriggers = buildKeyTriggersSection(availableAgents, availableSkills)
   const toolSelection = buildToolSelectionTable(availableAgents, availableTools, availableSkills)
   const exploreSection = buildExploreSection(availableAgents)
   const librarianSection = buildLibrarianSection(availableAgents)
-  const frontendSection = buildFrontendSection(availableAgents)
+  const categorySkillsGuide = buildCategorySkillsDelegationGuide(availableCategories, availableSkills)
   const delegationTable = buildDelegationTable(availableAgents)
   const oracleSection = buildOracleSection(availableAgents)
-  const hardBlocks = buildHardBlocksSection(availableAgents)
+  const hardBlocks = buildHardBlocksSection()
-  const antiPatterns = buildAntiPatternsSection(availableAgents)
+  const antiPatterns = buildAntiPatternsSection()
 
   const sections = [
     SISYPHUS_ROLE_SECTION,
@@ -567,7 +561,7 @@ function buildDynamicSisyphusPrompt(
     "",
     SISYPHUS_PHASE2B_PRE_IMPLEMENTATION,
     "",
-    frontendSection,
+    categorySkillsGuide,
     "",
     delegationTable,
     "",
@@ -608,18 +602,20 @@ export function createSisyphusAgent(
   model: string,
   availableAgents?: AvailableAgent[],
   availableToolNames?: string[],
-  availableSkills?: AvailableSkill[]
+  availableSkills?: AvailableSkill[],
+  availableCategories?: AvailableCategory[]
 ): AgentConfig {
   const tools = availableToolNames ? categorizeTools(availableToolNames) : []
   const skills = availableSkills ?? []
+  const categories = availableCategories ?? []
   const prompt = availableAgents
-    ? buildDynamicSisyphusPrompt(availableAgents, tools, skills)
+    ? buildDynamicSisyphusPrompt(availableAgents, tools, skills, categories)
-    : buildDynamicSisyphusPrompt([], tools, skills)
+    : buildDynamicSisyphusPrompt([], tools, skills, categories)
 
   const permission = { question: "allow", call_omo_agent: "deny" } as AgentConfig["permission"]
   const base = {
     description:
-      "Sisyphus - Powerful AI orchestrator from OhMyOpenCode. Plans obsessively with todos, assesses search complexity before exploration, delegates strategically to specialized agents. Uses explore for internal code (parallel-friendly), librarian only for external docs, and always delegates UI work to frontend engineer.",
+      "Sisyphus - Powerful AI orchestrator from OhMyOpenCode. Plans obsessively with todos, assesses search complexity before exploration, delegates strategically via category+skills combinations. Uses explore for internal code (parallel-friendly), librarian for external docs.",
     mode: "primary" as const,
     model,
     maxTokens: 64000,

src/agents/types.ts

@@ -61,12 +61,10 @@ export type BuiltinAgentName =
   | "oracle"
   | "librarian"
   | "explore"
-  | "frontend-ui-ux-engineer"
-  | "document-writer"
   | "multimodal-looker"
   | "Metis (Plan Consultant)"
   | "Momus (Plan Reviewer)"
-  | "orchestrator-sisyphus"
+  | "Atlas"
 
 export type OverridableAgentName =
   | "build"

src/agents/utils.test.ts

@@ -122,10 +122,8 @@ describe("buildAgent with category and skills", () => {
     // #when
     const agent = buildAgent(source["test-agent"], TEST_MODEL)
 
-    // #then - DEFAULT_CATEGORIES only has temperature, not model
+    // #then - category's built-in model is applied
-    // Model remains undefined since neither factory nor category provides it
+    expect(agent.model).toBe("google/gemini-3-pro-preview")
-    expect(agent.model).toBeUndefined()
-    expect(agent.temperature).toBe(0.7)
   })
 
   test("agent with category and existing model keeps existing model", () => {
@@ -142,9 +140,8 @@ describe("buildAgent with category and skills", () => {
     // #when
     const agent = buildAgent(source["test-agent"], TEST_MODEL)
 
-    // #then
+    // #then - explicit model takes precedence over category
     expect(agent.model).toBe("custom/model")
-    expect(agent.temperature).toBe(0.7)
   })
 
   test("agent with category inherits variant", () => {
@@ -247,9 +244,9 @@ describe("buildAgent with category and skills", () => {
     // #when
     const agent = buildAgent(source["test-agent"], TEST_MODEL)
 
-    // #then - DEFAULT_CATEGORIES["ultrabrain"] only has temperature, not model
+    // #then - category's built-in model and skills are applied
-    expect(agent.model).toBeUndefined()
+    expect(agent.model).toBe("openai/gpt-5.2-codex")
-    expect(agent.temperature).toBe(0.1)
+    expect(agent.variant).toBe("xhigh")
     expect(agent.prompt).toContain("Role: Designer-Turned-Developer")
     expect(agent.prompt).toContain("Task description")
   })

src/agents/utils.ts

@@ -5,16 +5,15 @@ import { createSisyphusAgent } from "./sisyphus"
 import { createOracleAgent, ORACLE_PROMPT_METADATA } from "./oracle"
 import { createLibrarianAgent, LIBRARIAN_PROMPT_METADATA } from "./librarian"
 import { createExploreAgent, EXPLORE_PROMPT_METADATA } from "./explore"
-import { createFrontendUiUxEngineerAgent, FRONTEND_PROMPT_METADATA } from "./frontend-ui-ux-engineer"
-import { createDocumentWriterAgent, DOCUMENT_WRITER_PROMPT_METADATA } from "./document-writer"
 import { createMultimodalLookerAgent, MULTIMODAL_LOOKER_PROMPT_METADATA } from "./multimodal-looker"
 import { createMetisAgent } from "./metis"
-import { createOrchestratorSisyphusAgent } from "./orchestrator-sisyphus"
+import { createAtlasAgent } from "./atlas"
 import { createMomusAgent } from "./momus"
-import type { AvailableAgent } from "./sisyphus-prompt-builder"
+import type { AvailableAgent, AvailableCategory, AvailableSkill } from "./dynamic-agent-prompt-builder"
 import { deepMerge } from "../shared"
-import { DEFAULT_CATEGORIES } from "../tools/delegate-task/constants"
+import { DEFAULT_CATEGORIES, CATEGORY_DESCRIPTIONS } from "../tools/delegate-task/constants"
 import { resolveMultipleSkills } from "../features/opencode-skill-loader/skill-content"
+import { createBuiltinSkills } from "../features/builtin-skills"
 
 type AgentSource = AgentFactory | AgentConfig
 
@@ -23,14 +22,12 @@ const agentSources: Record<BuiltinAgentName, AgentSource> = {
   oracle: createOracleAgent,
   librarian: createLibrarianAgent,
   explore: createExploreAgent,
-  "frontend-ui-ux-engineer": createFrontendUiUxEngineerAgent,
-  "document-writer": createDocumentWriterAgent,
   "multimodal-looker": createMultimodalLookerAgent,
   "Metis (Plan Consultant)": createMetisAgent,
   "Momus (Plan Reviewer)": createMomusAgent,
-  // Note: orchestrator-sisyphus is handled specially in createBuiltinAgents()
+  // Note: Atlas is handled specially in createBuiltinAgents()
   // because it needs OrchestratorContext, not just a model string
-  "orchestrator-sisyphus": createOrchestratorSisyphusAgent as unknown as AgentFactory,
+  Atlas: createAtlasAgent as unknown as AgentFactory,
 }
 
 /**
@@ -41,8 +38,6 @@ const agentMetadata: Partial<Record<BuiltinAgentName, AgentPromptMetadata>> = {
   oracle: ORACLE_PROMPT_METADATA,
   librarian: LIBRARIAN_PROMPT_METADATA,
   explore: EXPLORE_PROMPT_METADATA,
-  "frontend-ui-ux-engineer": FRONTEND_PROMPT_METADATA,
-  "document-writer": DOCUMENT_WRITER_PROMPT_METADATA,
   "multimodal-looker": MULTIMODAL_LOOKER_PROMPT_METADATA,
 }
 
@@ -155,11 +150,23 @@ export function createBuiltinAgents(
     ? { ...DEFAULT_CATEGORIES, ...categories }
     : DEFAULT_CATEGORIES
 
+  const availableCategories: AvailableCategory[] = Object.entries(mergedCategories).map(([name]) => ({
+    name,
+    description: CATEGORY_DESCRIPTIONS[name] ?? "General tasks",
+  }))
+
+  const builtinSkills = createBuiltinSkills()
+  const availableSkills: AvailableSkill[] = builtinSkills.map((skill) => ({
+    name: skill.name,
+    description: skill.description,
+    location: "plugin" as const,
+  }))
+
   for (const [name, source] of Object.entries(agentSources)) {
     const agentName = name as BuiltinAgentName
 
     if (agentName === "Sisyphus") continue
-    if (agentName === "orchestrator-sisyphus") continue
+    if (agentName === "Atlas") continue
     if (disabledAgents.includes(agentName)) continue
 
     const override = agentOverrides[agentName]
@@ -192,7 +199,13 @@ export function createBuiltinAgents(
     const sisyphusOverride = agentOverrides["Sisyphus"]
     const sisyphusModel = sisyphusOverride?.model ?? systemDefaultModel
 
-    let sisyphusConfig = createSisyphusAgent(sisyphusModel, availableAgents)
+    let sisyphusConfig = createSisyphusAgent(
+      sisyphusModel,
+      availableAgents,
+      undefined,
+      availableSkills,
+      availableCategories
+    )
 
     if (directory && sisyphusConfig.prompt) {
       const envContext = createEnvContext()
@@ -206,19 +219,21 @@ export function createBuiltinAgents(
     result["Sisyphus"] = sisyphusConfig
   }
 
-  if (!disabledAgents.includes("orchestrator-sisyphus")) {
+  if (!disabledAgents.includes("Atlas")) {
-    const orchestratorOverride = agentOverrides["orchestrator-sisyphus"]
+    const orchestratorOverride = agentOverrides["Atlas"]
     const orchestratorModel = orchestratorOverride?.model ?? systemDefaultModel
-    let orchestratorConfig = createOrchestratorSisyphusAgent({
+     let orchestratorConfig = createAtlasAgent({
        model: orchestratorModel,
        availableAgents,
+       availableSkills,
+       userCategories: categories,
      })
 
     if (orchestratorOverride) {
       orchestratorConfig = mergeAgentConfig(orchestratorConfig, orchestratorOverride)
     }
 
-    result["orchestrator-sisyphus"] = orchestratorConfig
+    result["Atlas"] = orchestratorConfig
   }
 
   return result

src/cli/config-manager.test.ts

@@ -200,60 +200,165 @@ describe("config-manager ANTIGRAVITY_PROVIDER_CONFIG", () => {
   })
 })
 
-describe("generateOmoConfig - v3 beta: no hardcoded models", () => {
+describe("generateOmoConfig - model fallback system", () => {
-  test("generates minimal config with only $schema", () => {
+  test("generates native sonnet models when Claude standard subscription", () => {
-    // #given any install config
+    // #given user has Claude standard subscription (not max20)
     const config: InstallConfig = {
       hasClaude: true,
       isMax20: false,
-      hasChatGPT: true,
+      hasOpenAI: false,
       hasGemini: false,
       hasCopilot: false,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: false,
     }
 
     // #when generating config
     const result = generateOmoConfig(config)
 
-    // #then should only contain $schema, no agents or categories
+    // #then should use native anthropic sonnet (cost-efficient for standard plan)
     expect(result.$schema).toBe("https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/master/assets/oh-my-opencode.schema.json")
-    expect(result.agents).toBeUndefined()
+    expect(result.agents).toBeDefined()
-    expect(result.categories).toBeUndefined()
+    expect((result.agents as Record<string, { model: string }>).Sisyphus.model).toBe("anthropic/claude-sonnet-4-5")
   })
 
-  test("does not include model fields regardless of provider config", () => {
+  test("generates native opus models when Claude max20 subscription", () => {
-    // #given user has multiple providers
+    // #given user has Claude max20 subscription
     const config: InstallConfig = {
       hasClaude: true,
       isMax20: true,
-      hasChatGPT: true,
+      hasOpenAI: false,
-      hasGemini: true,
+      hasGemini: false,
-      hasCopilot: true,
+      hasCopilot: false,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: false,
     }
 
     // #when generating config
     const result = generateOmoConfig(config)
 
-    // #then should not have agents or categories with model fields
+    // #then should use native anthropic opus (max power for max20 plan)
-    expect(result.agents).toBeUndefined()
+    expect((result.agents as Record<string, { model: string }>).Sisyphus.model).toBe("anthropic/claude-opus-4-5")
-    expect(result.categories).toBeUndefined()
   })
 
-  test("does not include model fields when no providers configured", () => {
+  test("uses github-copilot sonnet fallback when only copilot available", () => {
+    // #given user has only copilot (no max plan)
+    const config: InstallConfig = {
+      hasClaude: false,
+      isMax20: false,
+      hasOpenAI: false,
+      hasGemini: false,
+      hasCopilot: true,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: false,
+    }
+
+    // #when generating config
+    const result = generateOmoConfig(config)
+
+    // #then should use github-copilot sonnet models
+    expect((result.agents as Record<string, { model: string }>).Sisyphus.model).toBe("github-copilot/claude-sonnet-4.5")
+  })
+
+  test("uses ultimate fallback when no providers configured", () => {
     // #given user has no providers
     const config: InstallConfig = {
       hasClaude: false,
       isMax20: false,
-      hasChatGPT: false,
+      hasOpenAI: false,
       hasGemini: false,
       hasCopilot: false,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: false,
     }
 
     // #when generating config
     const result = generateOmoConfig(config)
 
-    // #then should still only contain $schema
+    // #then should use ultimate fallback for all agents
     expect(result.$schema).toBe("https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/master/assets/oh-my-opencode.schema.json")
-    expect(result.agents).toBeUndefined()
+    expect((result.agents as Record<string, { model: string }>).Sisyphus.model).toBe("opencode/glm-4.7-free")
-    expect(result.categories).toBeUndefined()
+  })
+
+  test("uses zai-coding-plan/glm-4.7 for librarian when Z.ai available", () => {
+    // #given user has Z.ai and Claude max20
+    const config: InstallConfig = {
+      hasClaude: true,
+      isMax20: true,
+      hasOpenAI: false,
+      hasGemini: false,
+      hasCopilot: false,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: true,
+    }
+
+    // #when generating config
+    const result = generateOmoConfig(config)
+
+    // #then librarian should use zai-coding-plan/glm-4.7
+    expect((result.agents as Record<string, { model: string }>).librarian.model).toBe("zai-coding-plan/glm-4.7")
+    // #then other agents should use native opus (max20 plan)
+    expect((result.agents as Record<string, { model: string }>).Sisyphus.model).toBe("anthropic/claude-opus-4-5")
+  })
+
+  test("uses native OpenAI models when only ChatGPT available", () => {
+    // #given user has only ChatGPT subscription
+    const config: InstallConfig = {
+      hasClaude: false,
+      isMax20: false,
+      hasOpenAI: true,
+      hasGemini: false,
+      hasCopilot: false,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: false,
+    }
+
+    // #when generating config
+    const result = generateOmoConfig(config)
+
+    // #then Sisyphus should use native OpenAI (fallback within native tier)
+    expect((result.agents as Record<string, { model: string }>).Sisyphus.model).toBe("openai/gpt-5.2")
+    // #then Oracle should use native OpenAI (primary for ultrabrain)
+    expect((result.agents as Record<string, { model: string }>).oracle.model).toBe("openai/gpt-5.2-codex")
+    // #then multimodal-looker should use native OpenAI (fallback within native tier)
+    expect((result.agents as Record<string, { model: string }>)["multimodal-looker"].model).toBe("openai/gpt-5.2")
+  })
+
+  test("uses haiku for explore when Claude max20", () => {
+    // #given user has Claude max20
+    const config: InstallConfig = {
+      hasClaude: true,
+      isMax20: true,
+      hasOpenAI: false,
+      hasGemini: false,
+      hasCopilot: false,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: false,
+    }
+
+    // #when generating config
+    const result = generateOmoConfig(config)
+
+    // #then explore should use haiku (max20 plan uses Claude quota)
+    expect((result.agents as Record<string, { model: string }>).explore.model).toBe("anthropic/claude-haiku-4-5")
+  })
+
+  test("uses grok-code for explore when not max20", () => {
+    // #given user has Claude but not max20
+    const config: InstallConfig = {
+      hasClaude: true,
+      isMax20: false,
+      hasOpenAI: false,
+      hasGemini: false,
+      hasCopilot: false,
+      hasOpencodeZen: false,
+      hasZaiCodingPlan: false,
+    }
+
+    // #when generating config
+    const result = generateOmoConfig(config)
+
+    // #then explore should use grok-code (preserve Claude quota)
+    expect((result.agents as Record<string, { model: string }>).explore.model).toBe("opencode/grok-code")
   })
 })

src/cli/config-manager.ts

@@ -6,6 +6,7 @@ import {
   type OpenCodeConfigPaths,
 } from "../shared"
 import type { ConfigMergeResult, DetectedConfig, InstallConfig } from "./types"
+import { generateModelConfig } from "./model-fallback"
 
 const OPENCODE_BINARIES = ["opencode", "opencode-desktop"] as const
 
@@ -306,14 +307,8 @@ function deepMerge<T extends Record<string, unknown>>(target: T, source: Partial
   return result
 }
 
-export function generateOmoConfig(_installConfig: InstallConfig): Record<string, unknown> {
+export function generateOmoConfig(installConfig: InstallConfig): Record<string, unknown> {
-  // v3 beta: No hardcoded model strings - users rely on their OpenCode configured model
+  return generateModelConfig(installConfig)
-  // Users who want specific models configure them explicitly after install
-  const config: Record<string, unknown> = {
-    $schema: "https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/master/assets/oh-my-opencode.schema.json",
-  }
-
-  return config
 }
 
 export function writeOmoConfig(installConfig: InstallConfig): ConfigMergeResult {
@@ -580,16 +575,40 @@ export function addProviderConfig(config: InstallConfig): ConfigMergeResult {
   }
 }
 
+function detectProvidersFromOmoConfig(): { hasOpenAI: boolean; hasOpencodeZen: boolean; hasZaiCodingPlan: boolean } {
+  const omoConfigPath = getOmoConfig()
+  if (!existsSync(omoConfigPath)) {
+    return { hasOpenAI: true, hasOpencodeZen: true, hasZaiCodingPlan: false }
+  }
+
+  try {
+    const content = readFileSync(omoConfigPath, "utf-8")
+    const omoConfig = parseJsonc<Record<string, unknown>>(content)
+    if (!omoConfig || typeof omoConfig !== "object") {
+      return { hasOpenAI: true, hasOpencodeZen: true, hasZaiCodingPlan: false }
+    }
+
+    const configStr = JSON.stringify(omoConfig)
+    const hasOpenAI = configStr.includes('"openai/')
+    const hasOpencodeZen = configStr.includes('"opencode/')
+    const hasZaiCodingPlan = configStr.includes('"zai-coding-plan/')
+
+    return { hasOpenAI, hasOpencodeZen, hasZaiCodingPlan }
+  } catch {
+    return { hasOpenAI: true, hasOpencodeZen: true, hasZaiCodingPlan: false }
+  }
+}
+
 export function detectCurrentConfig(): DetectedConfig {
-  // v3 beta: Since we no longer generate hardcoded model strings,
-  // detection only checks for plugin installation and Gemini auth plugin
   const result: DetectedConfig = {
     isInstalled: false,
     hasClaude: true,
     isMax20: true,
-    hasChatGPT: true,
+    hasOpenAI: true,
     hasGemini: false,
     hasCopilot: false,
+    hasOpencodeZen: true,
+    hasZaiCodingPlan: false,
   }
 
   const { format, path } = detectConfigFormat()
@@ -613,5 +632,10 @@ export function detectCurrentConfig(): DetectedConfig {
   // Gemini auth plugin detection still works via plugin presence
   result.hasGemini = plugins.some((p) => p.startsWith("opencode-antigravity-auth"))
 
+  const { hasOpenAI, hasOpencodeZen, hasZaiCodingPlan } = detectProvidersFromOmoConfig()
+  result.hasOpenAI = hasOpenAI
+  result.hasOpencodeZen = hasOpencodeZen
+  result.hasZaiCodingPlan = hasZaiCodingPlan
+
   return result
 }

src/cli/index.ts

@@ -24,28 +24,35 @@ program
   .description("Install and configure oh-my-opencode with interactive setup")
   .option("--no-tui", "Run in non-interactive mode (requires all options)")
   .option("--claude <value>", "Claude subscription: no, yes, max20")
-  .option("--chatgpt <value>", "ChatGPT subscription: no, yes")
+  .option("--openai <value>", "OpenAI/ChatGPT subscription: no, yes (default: no)")
   .option("--gemini <value>", "Gemini integration: no, yes")
   .option("--copilot <value>", "GitHub Copilot subscription: no, yes")
+  .option("--opencode-zen <value>", "OpenCode Zen access: no, yes (default: no)")
+  .option("--zai-coding-plan <value>", "Z.ai Coding Plan subscription: no, yes (default: no)")
   .option("--skip-auth", "Skip authentication setup hints")
   .addHelpText("after", `
 Examples:
   $ bunx oh-my-opencode install
-  $ bunx oh-my-opencode install --no-tui --claude=max20 --chatgpt=yes --gemini=yes --copilot=no
+  $ bunx oh-my-opencode install --no-tui --claude=max20 --openai=yes --gemini=yes --copilot=no
-  $ bunx oh-my-opencode install --no-tui --claude=no --chatgpt=no --gemini=no --copilot=yes
+  $ bunx oh-my-opencode install --no-tui --claude=no --gemini=no --copilot=yes --opencode-zen=yes
 
-Model Providers:
+Model Providers (Priority: Native > Copilot > OpenCode Zen > Z.ai):
-  Claude      Required for Sisyphus (main orchestrator) and Librarian agents
+  Claude        Native anthropic/ models (Opus, Sonnet, Haiku)
-  ChatGPT     Powers the Oracle agent for debugging and architecture
+  OpenAI        Native openai/ models (GPT-5.2 for Oracle)
-  Gemini      Powers frontend, documentation, and multimodal agents
+  Gemini        Native google/ models (Gemini 3 Pro, Flash)
+  Copilot       github-copilot/ models (fallback)
+  OpenCode Zen  opencode/ models (opencode/claude-opus-4-5, etc.)
+  Z.ai          zai-coding-plan/glm-4.7 (Librarian priority)
 `)
   .action(async (options) => {
     const args: InstallArgs = {
       tui: options.tui !== false,
       claude: options.claude,
-      chatgpt: options.chatgpt,
+      openai: options.openai,
       gemini: options.gemini,
       copilot: options.copilot,
+      opencodeZen: options.opencodeZen,
+      zaiCodingPlan: options.zaiCodingPlan,
       skipAuth: options.skipAuth ?? false,
     }
     const exitCode = await install(args)

src/cli/install.ts

@@ -10,6 +10,7 @@ import {
   addProviderConfig,
   detectCurrentConfig,
 } from "./config-manager"
+import { shouldShowChatGPTOnlyWarning } from "./model-fallback"
 import packageJson from "../../package.json" with { type: "json" }
 
 const VERSION = packageJson.version
@@ -39,19 +40,20 @@ function formatConfigSummary(config: InstallConfig): string {
 
   const claudeDetail = config.hasClaude ? (config.isMax20 ? "max20" : "standard") : undefined
   lines.push(formatProvider("Claude", config.hasClaude, claudeDetail))
-  lines.push(formatProvider("ChatGPT", config.hasChatGPT))
+  lines.push(formatProvider("OpenAI/ChatGPT", config.hasOpenAI, "GPT-5.2 for Oracle"))
   lines.push(formatProvider("Gemini", config.hasGemini))
-  lines.push(formatProvider("GitHub Copilot", config.hasCopilot, "fallback provider"))
+  lines.push(formatProvider("GitHub Copilot", config.hasCopilot, "fallback"))
+  lines.push(formatProvider("OpenCode Zen", config.hasOpencodeZen, "opencode/ models"))
+  lines.push(formatProvider("Z.ai Coding Plan", config.hasZaiCodingPlan, "Librarian: glm-4.7"))
 
   lines.push("")
   lines.push(color.dim("─".repeat(40)))
   lines.push("")
 
-  // v3 beta: No hardcoded models - agents use OpenCode's configured default model
+  lines.push(color.bold(color.white("Model Assignment")))
-  lines.push(color.bold(color.white("Agent Models")))
   lines.push("")
-  lines.push(`  ${SYMBOLS.info} Agents will use your OpenCode default model`)
+  lines.push(`  ${SYMBOLS.info} Models auto-configured based on provider priority`)
-  lines.push(`  ${SYMBOLS.bullet} Configure specific models in ${color.cyan("oh-my-opencode.json")} if needed`)
+  lines.push(`  ${SYMBOLS.bullet} Priority: Native > Copilot > OpenCode Zen > Z.ai`)
 
   return lines.join("\n")
 }
@@ -115,12 +117,6 @@ function validateNonTuiArgs(args: InstallArgs): { valid: boolean; errors: string
     errors.push(`Invalid --claude value: ${args.claude} (expected: no, yes, max20)`)
   }
 
-  if (args.chatgpt === undefined) {
-    errors.push("--chatgpt is required (values: no, yes)")
-  } else if (!["no", "yes"].includes(args.chatgpt)) {
-    errors.push(`Invalid --chatgpt value: ${args.chatgpt} (expected: no, yes)`)
-  }
-
   if (args.gemini === undefined) {
     errors.push("--gemini is required (values: no, yes)")
   } else if (!["no", "yes"].includes(args.gemini)) {
@@ -133,6 +129,18 @@ function validateNonTuiArgs(args: InstallArgs): { valid: boolean; errors: string
     errors.push(`Invalid --copilot value: ${args.copilot} (expected: no, yes)`)
   }
 
+  if (args.openai !== undefined && !["no", "yes"].includes(args.openai)) {
+    errors.push(`Invalid --openai value: ${args.openai} (expected: no, yes)`)
+  }
+
+  if (args.opencodeZen !== undefined && !["no", "yes"].includes(args.opencodeZen)) {
+    errors.push(`Invalid --opencode-zen value: ${args.opencodeZen} (expected: no, yes)`)
+  }
+
+  if (args.zaiCodingPlan !== undefined && !["no", "yes"].includes(args.zaiCodingPlan)) {
+    errors.push(`Invalid --zai-coding-plan value: ${args.zaiCodingPlan} (expected: no, yes)`)
+  }
+
   return { valid: errors.length === 0, errors }
 }
 
@@ -140,13 +148,15 @@ function argsToConfig(args: InstallArgs): InstallConfig {
   return {
     hasClaude: args.claude !== "no",
     isMax20: args.claude === "max20",
-    hasChatGPT: args.chatgpt === "yes",
+    hasOpenAI: args.openai === "yes",
     hasGemini: args.gemini === "yes",
     hasCopilot: args.copilot === "yes",
+    hasOpencodeZen: args.opencodeZen === "yes",
+    hasZaiCodingPlan: args.zaiCodingPlan === "yes",
   }
 }
 
-function detectedToInitialValues(detected: DetectedConfig): { claude: ClaudeSubscription; chatgpt: BooleanArg; gemini: BooleanArg; copilot: BooleanArg } {
+function detectedToInitialValues(detected: DetectedConfig): { claude: ClaudeSubscription; openai: BooleanArg; gemini: BooleanArg; copilot: BooleanArg; opencodeZen: BooleanArg; zaiCodingPlan: BooleanArg } {
   let claude: ClaudeSubscription = "no"
   if (detected.hasClaude) {
     claude = detected.isMax20 ? "max20" : "yes"
@@ -154,9 +164,11 @@ function detectedToInitialValues(detected: DetectedConfig): { claude: ClaudeSubs
 
   return {
     claude,
-    chatgpt: detected.hasChatGPT ? "yes" : "no",
+    openai: detected.hasOpenAI ? "yes" : "no",
     gemini: detected.hasGemini ? "yes" : "no",
     copilot: detected.hasCopilot ? "yes" : "no",
+    opencodeZen: detected.hasOpencodeZen ? "yes" : "no",
+    zaiCodingPlan: detected.hasZaiCodingPlan ? "yes" : "no",
   }
 }
 
@@ -178,16 +190,16 @@ async function runTuiMode(detected: DetectedConfig): Promise<InstallConfig | nul
     return null
   }
 
-  const chatgpt = await p.select({
+  const openai = await p.select({
-    message: "Do you have a ChatGPT Plus/Pro subscription?",
+    message: "Do you have an OpenAI/ChatGPT Plus subscription?",
     options: [
-      { value: "no" as const, label: "No", hint: "Oracle will use fallback model" },
+      { value: "no" as const, label: "No", hint: "Oracle will use fallback models" },
-      { value: "yes" as const, label: "Yes", hint: "GPT-5.2 for debugging and architecture" },
+      { value: "yes" as const, label: "Yes", hint: "GPT-5.2 for Oracle (high-IQ debugging)" },
     ],
-    initialValue: initial.chatgpt,
+    initialValue: initial.openai,
   })
 
-  if (p.isCancel(chatgpt)) {
+  if (p.isCancel(openai)) {
     p.cancel("Installation cancelled.")
     return null
   }
@@ -220,12 +232,42 @@ async function runTuiMode(detected: DetectedConfig): Promise<InstallConfig | nul
     return null
   }
 
+  const opencodeZen = await p.select({
+    message: "Do you have access to OpenCode Zen (opencode/ models)?",
+    options: [
+      { value: "no" as const, label: "No", hint: "Will use other configured providers" },
+      { value: "yes" as const, label: "Yes", hint: "opencode/claude-opus-4-5, opencode/gpt-5.2, etc." },
+    ],
+    initialValue: initial.opencodeZen,
+  })
+
+  if (p.isCancel(opencodeZen)) {
+    p.cancel("Installation cancelled.")
+    return null
+  }
+
+  const zaiCodingPlan = await p.select({
+    message: "Do you have a Z.ai Coding Plan subscription?",
+    options: [
+      { value: "no" as const, label: "No", hint: "Will use other configured providers" },
+      { value: "yes" as const, label: "Yes", hint: "zai-coding-plan/glm-4.7 for Librarian" },
+    ],
+    initialValue: initial.zaiCodingPlan,
+  })
+
+  if (p.isCancel(zaiCodingPlan)) {
+    p.cancel("Installation cancelled.")
+    return null
+  }
+
   return {
     hasClaude: claude !== "no",
     isMax20: claude === "max20",
-    hasChatGPT: chatgpt === "yes",
+    hasOpenAI: openai === "yes",
     hasGemini: gemini === "yes",
     hasCopilot: copilot === "yes",
+    hasOpencodeZen: opencodeZen === "yes",
+    hasZaiCodingPlan: zaiCodingPlan === "yes",
   }
 }
 
@@ -238,7 +280,7 @@ async function runNonTuiInstall(args: InstallArgs): Promise<number> {
       console.log(`  ${SYMBOLS.bullet} ${err}`)
     }
     console.log()
-    printInfo("Usage: bunx oh-my-opencode install --no-tui --claude=<no|yes|max20> --chatgpt=<no|yes> --gemini=<no|yes> --copilot=<no|yes>")
+    printInfo("Usage: bunx oh-my-opencode install --no-tui --claude=<no|yes|max20> --gemini=<no|yes> --copilot=<no|yes>")
     console.log()
     return 1
   }
@@ -264,7 +306,7 @@ async function runNonTuiInstall(args: InstallArgs): Promise<number> {
 
   if (isUpdate) {
     const initial = detectedToInitialValues(detected)
-    printInfo(`Current config: Claude=${initial.claude}, ChatGPT=${initial.chatgpt}, Gemini=${initial.gemini}`)
+    printInfo(`Current config: Claude=${initial.claude}, Gemini=${initial.gemini}`)
   }
 
   const config = argsToConfig(args)
@@ -307,7 +349,21 @@ async function runNonTuiInstall(args: InstallArgs): Promise<number> {
 
   printBox(formatConfigSummary(config), isUpdate ? "Updated Configuration" : "Installation Complete")
 
-  if (!config.hasClaude && !config.hasChatGPT && !config.hasGemini && !config.hasCopilot) {
+  if (!config.hasClaude) {
+    console.log()
+    console.log(color.bgRed(color.white(color.bold(" ⚠️  CRITICAL WARNING "))))
+    console.log()
+    console.log(color.red(color.bold("  Sisyphus agent is STRONGLY optimized for Claude Opus 4.5.")))
+    console.log(color.red("  Without Claude, you may experience significantly degraded performance:"))
+    console.log(color.dim("    • Reduced orchestration quality"))
+    console.log(color.dim("    • Weaker tool selection and delegation"))
+    console.log(color.dim("    • Less reliable task completion"))
+    console.log()
+    console.log(color.yellow("  Consider subscribing to Claude Pro/Max for the best experience."))
+    console.log()
+  }
+
+  if (!config.hasClaude && !config.hasOpenAI && !config.hasGemini && !config.hasCopilot && !config.hasOpencodeZen) {
     printWarning("No model providers configured. Using opencode/glm-4.7-free as fallback.")
   }
 
@@ -328,11 +384,10 @@ async function runNonTuiInstall(args: InstallArgs): Promise<number> {
   console.log(color.dim("oMoMoMoMo... Enjoy!"))
   console.log()
 
-  if ((config.hasClaude || config.hasChatGPT || config.hasGemini || config.hasCopilot) && !args.skipAuth) {
+  if ((config.hasClaude || config.hasGemini || config.hasCopilot) && !args.skipAuth) {
     printBox(
       `Run ${color.cyan("opencode auth login")} and select your provider:\n` +
       (config.hasClaude ? `  ${SYMBOLS.bullet} Anthropic ${color.gray("→ Claude Pro/Max")}\n` : "") +
-      (config.hasChatGPT ? `  ${SYMBOLS.bullet} OpenAI ${color.gray("→ ChatGPT Plus/Pro")}\n` : "") +
       (config.hasGemini ? `  ${SYMBOLS.bullet} Google ${color.gray("→ OAuth with Antigravity")}\n` : "") +
       (config.hasCopilot ? `  ${SYMBOLS.bullet} GitHub ${color.gray("→ Copilot")}` : ""),
       "🔐 Authenticate Your Providers"
@@ -354,7 +409,7 @@ export async function install(args: InstallArgs): Promise<number> {
 
   if (isUpdate) {
     const initial = detectedToInitialValues(detected)
-    p.log.info(`Existing configuration detected: Claude=${initial.claude}, ChatGPT=${initial.chatgpt}, Gemini=${initial.gemini}`)
+    p.log.info(`Existing configuration detected: Claude=${initial.claude}, Gemini=${initial.gemini}`)
   }
 
   const s = p.spinner()
@@ -413,7 +468,21 @@ export async function install(args: InstallArgs): Promise<number> {
   }
   s.stop(`Config written to ${color.cyan(omoResult.configPath)}`)
 
-  if (!config.hasClaude && !config.hasChatGPT && !config.hasGemini && !config.hasCopilot) {
+  if (!config.hasClaude) {
+    console.log()
+    console.log(color.bgRed(color.white(color.bold(" ⚠️  CRITICAL WARNING "))))
+    console.log()
+    console.log(color.red(color.bold("  Sisyphus agent is STRONGLY optimized for Claude Opus 4.5.")))
+    console.log(color.red("  Without Claude, you may experience significantly degraded performance:"))
+    console.log(color.dim("    • Reduced orchestration quality"))
+    console.log(color.dim("    • Weaker tool selection and delegation"))
+    console.log(color.dim("    • Less reliable task completion"))
+    console.log()
+    console.log(color.yellow("  Consider subscribing to Claude Pro/Max for the best experience."))
+    console.log()
+  }
+
+  if (!config.hasClaude && !config.hasOpenAI && !config.hasGemini && !config.hasCopilot && !config.hasOpencodeZen) {
     p.log.warn("No model providers configured. Using opencode/glm-4.7-free as fallback.")
   }
 
@@ -434,10 +503,9 @@ export async function install(args: InstallArgs): Promise<number> {
 
   p.outro(color.green("oMoMoMoMo... Enjoy!"))
 
-  if ((config.hasClaude || config.hasChatGPT || config.hasGemini || config.hasCopilot) && !args.skipAuth) {
+  if ((config.hasClaude || config.hasGemini || config.hasCopilot) && !args.skipAuth) {
     const providers: string[] = []
     if (config.hasClaude) providers.push(`Anthropic ${color.gray("→ Claude Pro/Max")}`)
-    if (config.hasChatGPT) providers.push(`OpenAI ${color.gray("→ ChatGPT Plus/Pro")}`)
     if (config.hasGemini) providers.push(`Google ${color.gray("→ OAuth with Antigravity")}`)
     if (config.hasCopilot) providers.push(`GitHub ${color.gray("→ Copilot")}`)
 

src/cli/model-fallback.ts

@@ -0,0 +1,246 @@
+import type { InstallConfig } from "./types"
+
+type NativeProvider = "claude" | "openai" | "gemini"
+
+type ModelCapability =
+  | "unspecified-high"
+  | "unspecified-low"
+  | "quick"
+  | "ultrabrain"
+  | "visual-engineering"
+  | "artistry"
+  | "writing"
+  | "glm"
+
+interface ProviderAvailability {
+  native: {
+    claude: boolean
+    openai: boolean
+    gemini: boolean
+  }
+  opencodeZen: boolean
+  copilot: boolean
+  zai: boolean
+  isMaxPlan: boolean
+}
+
+interface AgentConfig {
+  model: string
+  variant?: string
+}
+
+interface CategoryConfig {
+  model: string
+  variant?: string
+}
+
+export interface GeneratedOmoConfig {
+  $schema: string
+  agents?: Record<string, AgentConfig>
+  categories?: Record<string, CategoryConfig>
+  [key: string]: unknown
+}
+
+interface NativeFallbackEntry {
+  provider: NativeProvider
+  model: string
+}
+
+const NATIVE_FALLBACK_CHAINS: Record<ModelCapability, NativeFallbackEntry[]> = {
+  "unspecified-high": [
+    { provider: "claude", model: "anthropic/claude-opus-4-5" },
+    { provider: "openai", model: "openai/gpt-5.2" },
+    { provider: "gemini", model: "google/gemini-3-pro-preview" },
+  ],
+  "unspecified-low": [
+    { provider: "claude", model: "anthropic/claude-sonnet-4-5" },
+    { provider: "openai", model: "openai/gpt-5.2" },
+    { provider: "gemini", model: "google/gemini-3-flash-preview" },
+  ],
+  quick: [
+    { provider: "claude", model: "anthropic/claude-haiku-4-5" },
+    { provider: "openai", model: "openai/gpt-5.1-codex-mini" },
+    { provider: "gemini", model: "google/gemini-3-flash-preview" },
+  ],
+  ultrabrain: [
+    { provider: "openai", model: "openai/gpt-5.2-codex" },
+    { provider: "claude", model: "anthropic/claude-opus-4-5" },
+    { provider: "gemini", model: "google/gemini-3-pro-preview" },
+  ],
+  "visual-engineering": [
+    { provider: "gemini", model: "google/gemini-3-pro-preview" },
+    { provider: "openai", model: "openai/gpt-5.2" },
+    { provider: "claude", model: "anthropic/claude-sonnet-4-5" },
+  ],
+  artistry: [
+    { provider: "gemini", model: "google/gemini-3-pro-preview" },
+    { provider: "openai", model: "openai/gpt-5.2" },
+    { provider: "claude", model: "anthropic/claude-opus-4-5" },
+  ],
+  writing: [
+    { provider: "gemini", model: "google/gemini-3-flash-preview" },
+    { provider: "openai", model: "openai/gpt-5.2" },
+    { provider: "claude", model: "anthropic/claude-sonnet-4-5" },
+  ],
+  glm: [],
+}
+
+const OPENCODE_ZEN_MODELS: Record<ModelCapability, string> = {
+  "unspecified-high": "opencode/claude-opus-4-5",
+  "unspecified-low": "opencode/claude-sonnet-4-5",
+  quick: "opencode/claude-haiku-4-5",
+  ultrabrain: "opencode/gpt-5.2-codex",
+  "visual-engineering": "opencode/gemini-3-pro",
+  artistry: "opencode/gemini-3-pro",
+  writing: "opencode/gemini-3-flash",
+  glm: "opencode/glm-4.7-free",
+}
+
+const GITHUB_COPILOT_MODELS: Record<ModelCapability, string> = {
+  "unspecified-high": "github-copilot/claude-opus-4.5",
+  "unspecified-low": "github-copilot/claude-sonnet-4.5",
+  quick: "github-copilot/claude-haiku-4.5",
+  ultrabrain: "github-copilot/gpt-5.2-codex",
+  "visual-engineering": "github-copilot/gemini-3-pro-preview",
+  artistry: "github-copilot/gemini-3-pro-preview",
+  writing: "github-copilot/gemini-3-flash-preview",
+  glm: "github-copilot/gpt-5.2",
+}
+
+const ZAI_MODEL = "zai-coding-plan/glm-4.7"
+
+interface AgentRequirement {
+  capability: ModelCapability
+  variant?: string
+}
+
+const AGENT_REQUIREMENTS: Record<string, AgentRequirement> = {
+  Sisyphus: { capability: "unspecified-high" },
+  oracle: { capability: "ultrabrain", variant: "high" },
+  librarian: { capability: "glm" },
+  explore: { capability: "quick" },
+  "multimodal-looker": { capability: "visual-engineering" },
+  "Prometheus (Planner)": { capability: "unspecified-high" },
+  "Metis (Plan Consultant)": { capability: "unspecified-high" },
+  "Momus (Plan Reviewer)": { capability: "ultrabrain", variant: "medium" },
+  Atlas: { capability: "unspecified-high" },
+}
+
+interface CategoryRequirement {
+  capability: ModelCapability
+  variant?: string
+}
+
+const CATEGORY_REQUIREMENTS: Record<string, CategoryRequirement> = {
+  "visual-engineering": { capability: "visual-engineering" },
+  ultrabrain: { capability: "ultrabrain" },
+  artistry: { capability: "artistry", variant: "max" },
+  quick: { capability: "quick" },
+  "unspecified-low": { capability: "unspecified-low" },
+  "unspecified-high": { capability: "unspecified-high" },
+  writing: { capability: "writing" },
+}
+
+const ULTIMATE_FALLBACK = "opencode/glm-4.7-free"
+const SCHEMA_URL = "https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/master/assets/oh-my-opencode.schema.json"
+
+function toProviderAvailability(config: InstallConfig): ProviderAvailability {
+  return {
+    native: {
+      claude: config.hasClaude,
+      openai: config.hasOpenAI,
+      gemini: config.hasGemini,
+    },
+    opencodeZen: config.hasOpencodeZen,
+    copilot: config.hasCopilot,
+    zai: config.hasZaiCodingPlan,
+    isMaxPlan: config.isMax20,
+  }
+}
+
+function resolveModel(capability: ModelCapability, avail: ProviderAvailability): string {
+  const nativeChain = NATIVE_FALLBACK_CHAINS[capability]
+  for (const entry of nativeChain) {
+    if (avail.native[entry.provider]) {
+      return entry.model
+    }
+  }
+
+  if (avail.opencodeZen) {
+    return OPENCODE_ZEN_MODELS[capability]
+  }
+
+  if (avail.copilot) {
+    return GITHUB_COPILOT_MODELS[capability]
+  }
+
+  if (avail.zai) {
+    return ZAI_MODEL
+  }
+
+  return ULTIMATE_FALLBACK
+}
+
+function resolveClaudeCapability(avail: ProviderAvailability): ModelCapability {
+  return avail.isMaxPlan ? "unspecified-high" : "unspecified-low"
+}
+
+export function generateModelConfig(config: InstallConfig): GeneratedOmoConfig {
+  const avail = toProviderAvailability(config)
+  const hasAnyProvider =
+    avail.native.claude ||
+    avail.native.openai ||
+    avail.native.gemini ||
+    avail.opencodeZen ||
+    avail.copilot ||
+    avail.zai
+
+  if (!hasAnyProvider) {
+    return {
+      $schema: SCHEMA_URL,
+      agents: Object.fromEntries(
+        Object.keys(AGENT_REQUIREMENTS).map((role) => [role, { model: ULTIMATE_FALLBACK }])
+      ),
+      categories: Object.fromEntries(
+        Object.keys(CATEGORY_REQUIREMENTS).map((cat) => [cat, { model: ULTIMATE_FALLBACK }])
+      ),
+    }
+  }
+
+  const agents: Record<string, AgentConfig> = {}
+  const categories: Record<string, CategoryConfig> = {}
+
+  const claudeCapability = resolveClaudeCapability(avail)
+
+  for (const [role, req] of Object.entries(AGENT_REQUIREMENTS)) {
+    if (role === "librarian" && avail.zai) {
+      agents[role] = { model: ZAI_MODEL }
+    } else if (role === "explore") {
+      if (avail.native.claude && avail.isMaxPlan) {
+        agents[role] = { model: "anthropic/claude-haiku-4-5" }
+      } else {
+        agents[role] = { model: "opencode/grok-code" }
+      }
+    } else {
+      const capability = req.capability === "unspecified-high" ? claudeCapability : req.capability
+      const model = resolveModel(capability, avail)
+      agents[role] = req.variant ? { model, variant: req.variant } : { model }
+    }
+  }
+
+  for (const [cat, req] of Object.entries(CATEGORY_REQUIREMENTS)) {
+    const capability = req.capability === "unspecified-high" ? claudeCapability : req.capability
+    const model = resolveModel(capability, avail)
+    categories[cat] = req.variant ? { model, variant: req.variant } : { model }
+  }
+
+  return {
+    $schema: SCHEMA_URL,
+    agents,
+    categories,
+  }
+}
+
+export function shouldShowChatGPTOnlyWarning(config: InstallConfig): boolean {
+  return !config.hasClaude && !config.hasGemini && config.hasOpenAI
+}

src/cli/run/runner.ts

@@ -6,6 +6,8 @@ import { createEventState, processEvents, serializeError } from "./events"
 
 const POLL_INTERVAL_MS = 500
 const DEFAULT_TIMEOUT_MS = 0
+const SESSION_CREATE_MAX_RETRIES = 3
+const SESSION_CREATE_RETRY_DELAY_MS = 1000
 
 export async function run(options: RunOptions): Promise<number> {
   const {
@@ -45,13 +47,49 @@ export async function run(options: RunOptions): Promise<number> {
     })
 
     try {
+      // Retry session creation with exponential backoff
+      // Server might not be fully ready even after "listening" message
+      let sessionID: string | undefined
+      let lastError: unknown
+
+      for (let attempt = 1; attempt <= SESSION_CREATE_MAX_RETRIES; attempt++) {
         const sessionRes = await client.session.create({
           body: { title: "oh-my-opencode run" },
         })
 
-      const sessionID = sessionRes.data?.id
+        if (sessionRes.error) {
+          lastError = sessionRes.error
+          console.error(pc.yellow(`Session create attempt ${attempt}/${SESSION_CREATE_MAX_RETRIES} failed:`))
+          console.error(pc.dim(`  Error: ${serializeError(sessionRes.error)}`))
+
+          if (attempt < SESSION_CREATE_MAX_RETRIES) {
+            const delay = SESSION_CREATE_RETRY_DELAY_MS * attempt
+            console.log(pc.dim(`  Retrying in ${delay}ms...`))
+            await new Promise((resolve) => setTimeout(resolve, delay))
+            continue
+          }
+        }
+
+        sessionID = sessionRes.data?.id
+        if (sessionID) {
+          break
+        }
+
+        // No error but also no session ID - unexpected response
+        lastError = new Error(`Unexpected response: ${JSON.stringify(sessionRes, null, 2)}`)
+        console.error(pc.yellow(`Session create attempt ${attempt}/${SESSION_CREATE_MAX_RETRIES}: No session ID returned`))
+
+        if (attempt < SESSION_CREATE_MAX_RETRIES) {
+          const delay = SESSION_CREATE_RETRY_DELAY_MS * attempt
+          console.log(pc.dim(`  Retrying in ${delay}ms...`))
+          await new Promise((resolve) => setTimeout(resolve, delay))
+        }
+      }
+
       if (!sessionID) {
-        console.error(pc.red("Failed to create session"))
+        console.error(pc.red("Failed to create session after all retries"))
+        console.error(pc.dim(`Last error: ${serializeError(lastError)}`))
+        cleanup()
         return 1
       }
 

src/cli/types.ts

@@ -4,18 +4,22 @@ export type BooleanArg = "no" | "yes"
 export interface InstallArgs {
   tui: boolean
   claude?: ClaudeSubscription
-  chatgpt?: BooleanArg
+  openai?: BooleanArg
   gemini?: BooleanArg
   copilot?: BooleanArg
+  opencodeZen?: BooleanArg
+  zaiCodingPlan?: BooleanArg
   skipAuth?: boolean
 }
 
 export interface InstallConfig {
   hasClaude: boolean
   isMax20: boolean
-  hasChatGPT: boolean
+  hasOpenAI: boolean
   hasGemini: boolean
   hasCopilot: boolean
+  hasOpencodeZen: boolean
+  hasZaiCodingPlan: boolean
 }
 
 export interface ConfigMergeResult {
@@ -28,7 +32,9 @@ export interface DetectedConfig {
   isInstalled: boolean
   hasClaude: boolean
   isMax20: boolean
-  hasChatGPT: boolean
+  hasOpenAI: boolean
   hasGemini: boolean
   hasCopilot: boolean
+  hasOpencodeZen: boolean
+  hasZaiCodingPlan: boolean
 }

src/config/schema.test.ts

@@ -360,7 +360,7 @@ describe("CategoryConfigSchema", () => {
 describe("BuiltinCategoryNameSchema", () => {
   test("accepts all builtin category names", () => {
     // #given
-    const categories = ["visual-engineering", "ultrabrain", "artistry", "quick", "most-capable", "writing", "general"]
+    const categories = ["visual-engineering", "ultrabrain", "artistry", "quick", "unspecified-low", "unspecified-high", "writing"]
 
     // #when / #then
     for (const cat of categories) {

src/config/schema.ts

@@ -21,12 +21,10 @@ export const BuiltinAgentNameSchema = z.enum([
   "oracle",
   "librarian",
   "explore",
-  "frontend-ui-ux-engineer",
-  "document-writer",
   "multimodal-looker",
   "Metis (Plan Consultant)",
   "Momus (Plan Reviewer)",
-  "orchestrator-sisyphus",
+  "Atlas",
 ])
 
 export const BuiltinSkillNameSchema = z.enum([
@@ -47,10 +45,8 @@ export const OverridableAgentNameSchema = z.enum([
   "oracle",
   "librarian",
   "explore",
-  "frontend-ui-ux-engineer",
-  "document-writer",
   "multimodal-looker",
-  "orchestrator-sisyphus",
+  "Atlas",
 ])
 
 export const AgentNameSchema = BuiltinAgentNameSchema
@@ -87,7 +83,7 @@ export const HookNameSchema = z.enum([
   "delegate-task-retry",
   "prometheus-md-only",
   "start-work",
-  "sisyphus-orchestrator",
+  "atlas",
 ])
 
 export const BuiltinCommandNameSchema = z.enum([
@@ -130,10 +126,8 @@ export const AgentOverridesSchema = z.object({
   oracle: AgentOverrideConfigSchema.optional(),
   librarian: AgentOverrideConfigSchema.optional(),
   explore: AgentOverrideConfigSchema.optional(),
-  "frontend-ui-ux-engineer": AgentOverrideConfigSchema.optional(),
-  "document-writer": AgentOverrideConfigSchema.optional(),
   "multimodal-looker": AgentOverrideConfigSchema.optional(),
-  "orchestrator-sisyphus": AgentOverrideConfigSchema.optional(),
+  Atlas: AgentOverrideConfigSchema.optional(),
 })
 
 export const ClaudeCodeConfigSchema = z.object({
@@ -167,6 +161,8 @@ export const CategoryConfigSchema = z.object({
   textVerbosity: z.enum(["low", "medium", "high"]).optional(),
   tools: z.record(z.string(), z.boolean()).optional(),
   prompt_append: z.string().optional(),
+  /** Mark agent as unstable - forces background mode for monitoring. Auto-enabled for gemini models. */
+  is_unstable_agent: z.boolean().optional(),
 })
 
 export const BuiltinCategoryNameSchema = z.enum([
@@ -174,9 +170,9 @@ export const BuiltinCategoryNameSchema = z.enum([
   "ultrabrain",
   "artistry",
   "quick",
-  "most-capable",
+  "unspecified-low",
+  "unspecified-high",
   "writing",
-  "general",
 ])
 
 export const CategoriesConfigSchema = z.record(z.string(), CategoryConfigSchema)

src/features/AGENTS.md

@@ -36,7 +36,7 @@ features/
 | Type | Priority (highest first) |
 |------|--------------------------|
 | Commands | `.opencode/command/` > `~/.config/opencode/command/` > `.claude/commands/` > `~/.claude/commands/` |
-| Skills | `.opencode/skill/` > `~/.config/opencode/skill/` > `.claude/skills/` > `~/.claude/skills/` |
+| Skills | `.opencode/skills/` > `~/.config/opencode/skills/` > `.claude/skills/` > `~/.claude/skills/` |
 | Agents | `.claude/agents/` > `~/.claude/agents/` |
 | MCPs | `.claude/.mcp.json` > `.mcp.json` > `~/.claude/.mcp.json` |
 

src/features/builtin-commands/commands.ts

@@ -55,7 +55,7 @@ ${REFACTOR_TEMPLATE}
   },
   "start-work": {
     description: "(builtin) Start Sisyphus work session from Prometheus plan",
-    agent: "orchestrator-sisyphus",
+    agent: "Atlas",
     template: `<command-instruction>
 ${START_WORK_TEMPLATE}
 </command-instruction>

src/features/builtin-commands/templates/init-deep.ts

@@ -236,11 +236,11 @@ AGENTS_LOCATIONS = [
 
 ### Subdirectory AGENTS.md (Parallel)
 
-Launch document-writer agents for each location:
+Launch writing tasks for each location:
 
 \`\`\`
 for loc in AGENTS_LOCATIONS (except root):
-  delegate_task(agent="document-writer", prompt=\\\`
+  delegate_task(category="writing", prompt=\\\`
     Generate AGENTS.md for: \${loc.path}
     - Reason: \${loc.reason}
     - 30-80 lines max

src/features/opencode-skill-loader/async-loader.test.ts

@@ -5,7 +5,7 @@ import { tmpdir } from "os"
 import type { LoadedSkill } from "./types"
 
 const TEST_DIR = join(tmpdir(), "async-loader-test-" + Date.now())
-const SKILLS_DIR = join(TEST_DIR, ".opencode", "skill")
+const SKILLS_DIR = join(TEST_DIR, ".opencode", "skills")
 
 function createTestSkill(name: string, content: string, mcpJson?: object): string {
   const skillDir = join(SKILLS_DIR, name)

src/features/opencode-skill-loader/loader.test.ts

@@ -4,7 +4,7 @@ import { join } from "path"
 import { tmpdir } from "os"
 
 const TEST_DIR = join(tmpdir(), "skill-loader-test-" + Date.now())
-const SKILLS_DIR = join(TEST_DIR, ".opencode", "skill")
+const SKILLS_DIR = join(TEST_DIR, ".opencode", "skills")
 
 function createTestSkill(name: string, content: string, mcpJson?: object): string {
   const skillDir = join(SKILLS_DIR, name)

src/features/opencode-skill-loader/loader.ts

@@ -1,11 +1,11 @@
 import { promises as fs } from "fs"
 import { join, basename } from "path"
-import { homedir } from "os"
 import yaml from "js-yaml"
 import { parseFrontmatter } from "../../shared/frontmatter"
 import { sanitizeModelField } from "../../shared/model-sanitizer"
 import { resolveSymlinkAsync, isMarkdownFile } from "../../shared/file-utils"
 import { getClaudeConfigDir } from "../../shared"
+import { getOpenCodeConfigDir } from "../../shared/opencode-config-dir"
 import type { CommandDefinition } from "../claude-code-command-loader/types"
 import type { SkillScope, SkillMetadata, LoadedSkill, LazyContentLoader } from "./types"
 import type { SkillMcpConfig } from "../skill-mcp-manager/types"
@@ -187,13 +187,14 @@ export async function loadProjectSkills(): Promise<Record<string, CommandDefinit
 }
 
 export async function loadOpencodeGlobalSkills(): Promise<Record<string, CommandDefinition>> {
-  const opencodeSkillsDir = join(homedir(), ".config", "opencode", "skill")
+  const configDir = getOpenCodeConfigDir({ binary: "opencode" })
+  const opencodeSkillsDir = join(configDir, "skills")
   const skills = await loadSkillsFromDir(opencodeSkillsDir, "opencode")
   return skillsToRecord(skills)
 }
 
 export async function loadOpencodeProjectSkills(): Promise<Record<string, CommandDefinition>> {
-  const opencodeProjectDir = join(process.cwd(), ".opencode", "skill")
+  const opencodeProjectDir = join(process.cwd(), ".opencode", "skills")
   const skills = await loadSkillsFromDir(opencodeProjectDir, "opencode-project")
   return skillsToRecord(skills)
 }
@@ -249,11 +250,12 @@ export async function discoverProjectClaudeSkills(): Promise<LoadedSkill[]> {
 }
 
 export async function discoverOpencodeGlobalSkills(): Promise<LoadedSkill[]> {
-  const opencodeSkillsDir = join(homedir(), ".config", "opencode", "skill")
+  const configDir = getOpenCodeConfigDir({ binary: "opencode" })
+  const opencodeSkillsDir = join(configDir, "skills")
   return loadSkillsFromDir(opencodeSkillsDir, "opencode")
 }
 
 export async function discoverOpencodeProjectSkills(): Promise<LoadedSkill[]> {
-  const opencodeProjectDir = join(process.cwd(), ".opencode", "skill")
+  const opencodeProjectDir = join(process.cwd(), ".opencode", "skills")
   return loadSkillsFromDir(opencodeProjectDir, "opencode-project")
 }

src/hooks/AGENTS.md

@@ -8,7 +8,7 @@
 
 ```
 hooks/
-├── sisyphus-orchestrator/      # Main orchestration & delegation (771 lines)
+├── atlas/                      # Main orchestration & delegation (771 lines)
 ├── anthropic-context-window-limit-recovery/  # Auto-summarize at token limit
 ├── todo-continuation-enforcer.ts # Force TODO completion
 ├── ralph-loop/                 # Self-referential dev loop until done

src/hooks/atlas/index.test.ts

@@ -2,7 +2,7 @@ import { describe, expect, test, beforeEach, afterEach, mock } from "bun:test"
 import { existsSync, mkdirSync, rmSync, writeFileSync } from "node:fs"
 import { join } from "node:path"
 import { tmpdir } from "node:os"
-import { createSisyphusOrchestratorHook } from "./index"
+import { createAtlasHook } from "./index"
 import {
   writeBoulderState,
   clearBoulderState,
@@ -12,8 +12,8 @@ import type { BoulderState } from "../../features/boulder-state"
 
 import { MESSAGE_STORAGE } from "../../features/hook-message-injector"
 
-describe("sisyphus-orchestrator hook", () => {
+describe("atlas hook", () => {
-  const TEST_DIR = join(tmpdir(), "sisyphus-orchestrator-test-" + Date.now())
+   const TEST_DIR = join(tmpdir(), "atlas-test-" + Date.now())
   const SISYPHUS_DIR = join(TEST_DIR, ".sisyphus")
 
   function createMockPluginInput(overrides?: { promptMock?: ReturnType<typeof mock> }) {
@@ -26,7 +26,7 @@ describe("sisyphus-orchestrator hook", () => {
         },
       },
       _promptMock: promptMock,
-    } as unknown as Parameters<typeof createSisyphusOrchestratorHook>[0] & { _promptMock: ReturnType<typeof mock> }
+    } as unknown as Parameters<typeof createAtlasHook>[0] & { _promptMock: ReturnType<typeof mock> }
   }
 
   function setupMessageStorage(sessionID: string, agent: string): void {
@@ -68,7 +68,7 @@ describe("sisyphus-orchestrator hook", () => {
   describe("tool.execute.after handler", () => {
     test("should ignore non-delegate_task tools", async () => {
       // #given - hook and non-delegate_task tool
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Test Tool",
         output: "Original output",
@@ -85,8 +85,8 @@ describe("sisyphus-orchestrator hook", () => {
       expect(output.output).toBe("Original output")
     })
 
-    test("should not transform when caller is not orchestrator-sisyphus", async () => {
+     test("should not transform when caller is not Atlas", async () => {
-      // #given - boulder state exists but caller agent in message storage is not orchestrator
+       // #given - boulder state exists but caller agent in message storage is not Atlas
        const sessionID = "session-non-orchestrator-test"
        setupMessageStorage(sessionID, "other-agent")
       
@@ -101,7 +101,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
       writeBoulderState(TEST_DIR, state)
 
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Task completed successfully",
@@ -120,12 +120,12 @@ describe("sisyphus-orchestrator hook", () => {
       cleanupMessageStorage(sessionID)
     })
 
-    test("should append standalone verification when no boulder state but caller is orchestrator", async () => {
+     test("should append standalone verification when no boulder state but caller is Atlas", async () => {
-      // #given - no boulder state, but caller is orchestrator
+       // #given - no boulder state, but caller is Atlas
        const sessionID = "session-no-boulder-test"
-      setupMessageStorage(sessionID, "orchestrator-sisyphus")
+       setupMessageStorage(sessionID, "Atlas")
       
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Task completed successfully",
@@ -146,10 +146,10 @@ describe("sisyphus-orchestrator hook", () => {
       cleanupMessageStorage(sessionID)
     })
 
-    test("should transform output when caller is orchestrator-sisyphus with boulder state", async () => {
+     test("should transform output when caller is Atlas with boulder state", async () => {
-      // #given - orchestrator-sisyphus caller with boulder state
+       // #given - Atlas caller with boulder state
        const sessionID = "session-transform-test"
-      setupMessageStorage(sessionID, "orchestrator-sisyphus")
+       setupMessageStorage(sessionID, "Atlas")
       
       const planPath = join(TEST_DIR, "test-plan.md")
       writeFileSync(planPath, "# Plan\n- [ ] Task 1\n- [x] Task 2")
@@ -162,7 +162,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
       writeBoulderState(TEST_DIR, state)
 
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Task completed successfully",
@@ -186,9 +186,9 @@ describe("sisyphus-orchestrator hook", () => {
     })
 
      test("should still transform when plan is complete (shows progress)", async () => {
-      // #given - boulder state with complete plan, orchestrator caller
+       // #given - boulder state with complete plan, Atlas caller
        const sessionID = "session-complete-plan-test"
-      setupMessageStorage(sessionID, "orchestrator-sisyphus")
+       setupMessageStorage(sessionID, "Atlas")
       
       const planPath = join(TEST_DIR, "complete-plan.md")
       writeFileSync(planPath, "# Plan\n- [x] Task 1\n- [x] Task 2")
@@ -201,7 +201,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
       writeBoulderState(TEST_DIR, state)
 
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Original output",
@@ -223,9 +223,9 @@ describe("sisyphus-orchestrator hook", () => {
     })
 
      test("should append session ID to boulder state if not present", async () => {
-      // #given - boulder state without session-append-test, orchestrator caller
+       // #given - boulder state without session-append-test, Atlas caller
        const sessionID = "session-append-test"
-      setupMessageStorage(sessionID, "orchestrator-sisyphus")
+       setupMessageStorage(sessionID, "Atlas")
       
       const planPath = join(TEST_DIR, "test-plan.md")
       writeFileSync(planPath, "# Plan\n- [ ] Task 1")
@@ -238,7 +238,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
       writeBoulderState(TEST_DIR, state)
 
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Task output",
@@ -259,9 +259,9 @@ describe("sisyphus-orchestrator hook", () => {
     })
 
      test("should not duplicate existing session ID", async () => {
-      // #given - boulder state already has session-dup-test, orchestrator caller
+       // #given - boulder state already has session-dup-test, Atlas caller
        const sessionID = "session-dup-test"
-      setupMessageStorage(sessionID, "orchestrator-sisyphus")
+       setupMessageStorage(sessionID, "Atlas")
       
       const planPath = join(TEST_DIR, "test-plan.md")
       writeFileSync(planPath, "# Plan\n- [ ] Task 1")
@@ -274,7 +274,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
       writeBoulderState(TEST_DIR, state)
 
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Task output",
@@ -296,9 +296,9 @@ describe("sisyphus-orchestrator hook", () => {
     })
 
      test("should include boulder.json path and notepad path in transformed output", async () => {
-      // #given - boulder state, orchestrator caller
+       // #given - boulder state, Atlas caller
        const sessionID = "session-path-test"
-      setupMessageStorage(sessionID, "orchestrator-sisyphus")
+       setupMessageStorage(sessionID, "Atlas")
       
       const planPath = join(TEST_DIR, "my-feature.md")
       writeFileSync(planPath, "# Plan\n- [ ] Task 1\n- [ ] Task 2\n- [x] Task 3")
@@ -311,7 +311,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
       writeBoulderState(TEST_DIR, state)
 
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Task completed",
@@ -333,9 +333,9 @@ describe("sisyphus-orchestrator hook", () => {
     })
 
      test("should include resume and checkbox instructions in reminder", async () => {
-      // #given - boulder state, orchestrator caller
+       // #given - boulder state, Atlas caller
        const sessionID = "session-resume-test"
-      setupMessageStorage(sessionID, "orchestrator-sisyphus")
+       setupMessageStorage(sessionID, "Atlas")
       
       const planPath = join(TEST_DIR, "test-plan.md")
       writeFileSync(planPath, "# Plan\n- [ ] Task 1")
@@ -348,7 +348,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
       writeBoulderState(TEST_DIR, state)
 
-      const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+      const hook = createAtlasHook(createMockPluginInput())
       const output = {
         title: "Sisyphus Task",
         output: "Task completed",
@@ -373,7 +373,7 @@ describe("sisyphus-orchestrator hook", () => {
       const ORCHESTRATOR_SESSION = "orchestrator-write-test"
 
        beforeEach(() => {
-        setupMessageStorage(ORCHESTRATOR_SESSION, "orchestrator-sisyphus")
+         setupMessageStorage(ORCHESTRATOR_SESSION, "Atlas")
        })
 
       afterEach(() => {
@@ -382,7 +382,7 @@ describe("sisyphus-orchestrator hook", () => {
 
       test("should append delegation reminder when orchestrator writes outside .sisyphus/", async () => {
         // #given
-        const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+        const hook = createAtlasHook(createMockPluginInput())
         const output = {
           title: "Write",
           output: "File written successfully",
@@ -403,7 +403,7 @@ describe("sisyphus-orchestrator hook", () => {
 
       test("should append delegation reminder when orchestrator edits outside .sisyphus/", async () => {
         // #given
-        const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+        const hook = createAtlasHook(createMockPluginInput())
         const output = {
           title: "Edit",
           output: "File edited successfully",
@@ -422,7 +422,7 @@ describe("sisyphus-orchestrator hook", () => {
 
       test("should NOT append reminder when orchestrator writes inside .sisyphus/", async () => {
         // #given
-        const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+        const hook = createAtlasHook(createMockPluginInput())
         const originalOutput = "File written successfully"
         const output = {
           title: "Write",
@@ -446,7 +446,7 @@ describe("sisyphus-orchestrator hook", () => {
         const nonOrchestratorSession = "non-orchestrator-session"
         setupMessageStorage(nonOrchestratorSession, "Sisyphus-Junior")
         
-        const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+        const hook = createAtlasHook(createMockPluginInput())
         const originalOutput = "File written successfully"
         const output = {
           title: "Write",
@@ -469,7 +469,7 @@ describe("sisyphus-orchestrator hook", () => {
 
       test("should NOT append reminder for read-only tools", async () => {
         // #given
-        const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+        const hook = createAtlasHook(createMockPluginInput())
         const originalOutput = "File content"
         const output = {
           title: "Read",
@@ -489,7 +489,7 @@ describe("sisyphus-orchestrator hook", () => {
 
       test("should handle missing filePath gracefully", async () => {
         // #given
-        const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+        const hook = createAtlasHook(createMockPluginInput())
         const originalOutput = "File written successfully"
         const output = {
           title: "Write",
@@ -510,7 +510,7 @@ describe("sisyphus-orchestrator hook", () => {
       describe("cross-platform path validation (Windows support)", () => {
         test("should NOT append reminder when orchestrator writes inside .sisyphus\\ (Windows backslash)", async () => {
           // #given
-          const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+          const hook = createAtlasHook(createMockPluginInput())
           const originalOutput = "File written successfully"
           const output = {
             title: "Write",
@@ -531,7 +531,7 @@ describe("sisyphus-orchestrator hook", () => {
 
         test("should NOT append reminder when orchestrator writes inside .sisyphus with mixed separators", async () => {
           // #given
-          const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+          const hook = createAtlasHook(createMockPluginInput())
           const originalOutput = "File written successfully"
           const output = {
             title: "Write",
@@ -552,7 +552,7 @@ describe("sisyphus-orchestrator hook", () => {
 
         test("should NOT append reminder for absolute Windows path inside .sisyphus\\", async () => {
           // #given
-          const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+          const hook = createAtlasHook(createMockPluginInput())
           const originalOutput = "File written successfully"
           const output = {
             title: "Write",
@@ -573,7 +573,7 @@ describe("sisyphus-orchestrator hook", () => {
 
         test("should append reminder for Windows path outside .sisyphus\\", async () => {
           // #given
-          const hook = createSisyphusOrchestratorHook(createMockPluginInput())
+          const hook = createAtlasHook(createMockPluginInput())
           const output = {
             title: "Write",
             output: "File written successfully",
@@ -601,7 +601,7 @@ describe("sisyphus-orchestrator hook", () => {
          getMainSessionID: () => MAIN_SESSION_ID,
          subagentSessions: new Set<string>(),
        }))
-      setupMessageStorage(MAIN_SESSION_ID, "orchestrator-sisyphus")
+       setupMessageStorage(MAIN_SESSION_ID, "Atlas")
      })
 
     afterEach(() => {
@@ -622,7 +622,7 @@ describe("sisyphus-orchestrator hook", () => {
       writeBoulderState(TEST_DIR, state)
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when
       await hook.handler({
@@ -643,7 +643,7 @@ describe("sisyphus-orchestrator hook", () => {
     test("should not inject when no boulder state exists", async () => {
       // #given - no boulder state
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when
       await hook.handler({
@@ -671,7 +671,7 @@ describe("sisyphus-orchestrator hook", () => {
       writeBoulderState(TEST_DIR, state)
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when
       await hook.handler({
@@ -699,7 +699,7 @@ describe("sisyphus-orchestrator hook", () => {
       writeBoulderState(TEST_DIR, state)
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when - send abort error then idle
       await hook.handler({
@@ -740,7 +740,7 @@ describe("sisyphus-orchestrator hook", () => {
       }
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput, {
+      const hook = createAtlasHook(mockInput, {
         directory: TEST_DIR,
         backgroundManager: mockBackgroundManager as any,
       })
@@ -771,7 +771,7 @@ describe("sisyphus-orchestrator hook", () => {
       writeBoulderState(TEST_DIR, state)
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when - abort error, then message update, then idle
       await hook.handler({
@@ -814,7 +814,7 @@ describe("sisyphus-orchestrator hook", () => {
       writeBoulderState(TEST_DIR, state)
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when
       await hook.handler({
@@ -830,8 +830,8 @@ describe("sisyphus-orchestrator hook", () => {
       expect(callArgs.body.parts[0].text).toContain("2 remaining")
     })
 
-    test("should not inject when last agent is not orchestrator-sisyphus", async () => {
+     test("should not inject when last agent is not Atlas", async () => {
-      // #given - boulder state with incomplete plan, but last agent is NOT orchestrator-sisyphus
+       // #given - boulder state with incomplete plan, but last agent is NOT Atlas
        const planPath = join(TEST_DIR, "test-plan.md")
        writeFileSync(planPath, "# Plan\n- [ ] Task 1\n- [ ] Task 2")
 
@@ -843,12 +843,12 @@ describe("sisyphus-orchestrator hook", () => {
        }
        writeBoulderState(TEST_DIR, state)
 
-      // #given - last agent is NOT orchestrator-sisyphus
+       // #given - last agent is NOT Atlas
        cleanupMessageStorage(MAIN_SESSION_ID)
        setupMessageStorage(MAIN_SESSION_ID, "Sisyphus")
 
        const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+       const hook = createAtlasHook(mockInput)
 
        // #when
        await hook.handler({
@@ -858,7 +858,7 @@ describe("sisyphus-orchestrator hook", () => {
          },
        })
 
-      // #then - should NOT call prompt because agent is not orchestrator-sisyphus
+       // #then - should NOT call prompt because agent is not Atlas
        expect(mockInput._promptMock).not.toHaveBeenCalled()
      })
 
@@ -876,7 +876,7 @@ describe("sisyphus-orchestrator hook", () => {
       writeBoulderState(TEST_DIR, state)
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when - fire multiple idle events in rapid succession (simulating infinite loop bug)
       await hook.handler({
@@ -916,7 +916,7 @@ describe("sisyphus-orchestrator hook", () => {
       writeBoulderState(TEST_DIR, state)
 
       const mockInput = createMockPluginInput()
-      const hook = createSisyphusOrchestratorHook(mockInput)
+      const hook = createAtlasHook(mockInput)
 
       // #when - create abort state then delete
       await hook.handler({

src/hooks/atlas/index.ts

@@ -13,7 +13,7 @@ import { log } from "../../shared/logger"
 import { createSystemDirective, SYSTEM_DIRECTIVE_PREFIX, SystemDirectiveTypes } from "../../shared/system-directive"
 import type { BackgroundManager } from "../../features/background-agent"
 
-export const HOOK_NAME = "sisyphus-orchestrator"
+export const HOOK_NAME = "atlas"
 
 /**
  * Cross-platform check if a path is inside .sisyphus/ directory.
@@ -111,7 +111,7 @@ const ORCHESTRATOR_DELEGATION_REQUIRED = `
 
 **STOP. YOU ARE VIOLATING ORCHESTRATOR PROTOCOL.**
 
-You (orchestrator-sisyphus) are attempting to directly modify a file outside \`.sisyphus/\`.
+You (Atlas) are attempting to directly modify a file outside \`.sisyphus/\`.
 
 **Path attempted:** $FILE_PATH
 
@@ -397,8 +397,8 @@ function isCallerOrchestrator(sessionID?: string): boolean {
    const messageDir = getMessageDir(sessionID)
    if (!messageDir) return false
    const nearest = findNearestMessageWithFields(messageDir)
-  return nearest?.agent === "orchestrator-sisyphus"
+   return nearest?.agent === "Atlas"
-}
+ }
 
 interface SessionState {
   lastEventWasAbortError?: boolean
@@ -407,7 +407,7 @@ interface SessionState {
 
 const CONTINUATION_COOLDOWN_MS = 5000
 
-export interface SisyphusOrchestratorHookOptions {
+export interface AtlasHookOptions {
   directory: string
   backgroundManager?: BackgroundManager
 }
@@ -433,9 +433,9 @@ function isAbortError(error: unknown): boolean {
   return false
 }
 
-export function createSisyphusOrchestratorHook(
+export function createAtlasHook(
   ctx: PluginInput,
-  options?: SisyphusOrchestratorHookOptions
+  options?: AtlasHookOptions
 ) {
   const backgroundManager = options?.backgroundManager
   const sessions = new Map<string, SessionState>()
@@ -496,7 +496,7 @@ export function createSisyphusOrchestratorHook(
        await ctx.client.session.prompt({
          path: { id: sessionID },
          body: {
-          agent: "orchestrator-sisyphus",
+            agent: "Atlas",
            ...(model !== undefined ? { model } : {}),
            parts: [{ type: "text", text: prompt }],
          },
@@ -569,7 +569,7 @@ export function createSisyphusOrchestratorHook(
         }
 
         if (!isCallerOrchestrator(sessionID)) {
-          log(`[${HOOK_NAME}] Skipped: last agent is not orchestrator-sisyphus`, { sessionID })
+          log(`[${HOOK_NAME}] Skipped: last agent is not Atlas`, { sessionID })
           return
         }
 

src/hooks/delegate-task-retry/index.ts

@@ -108,7 +108,7 @@ Example of CORRECT call:
 delegate_task(
   description="Task description",
   prompt="Detailed prompt...",
-  category="general",  // OR subagent_type="explore"
+  category="unspecified-low",  // OR subagent_type="explore"
   run_in_background=false,
   skills=[]
 )

src/hooks/index.ts

@@ -28,5 +28,5 @@ export { createEditErrorRecoveryHook } from "./edit-error-recovery";
 export { createPrometheusMdOnlyHook } from "./prometheus-md-only";
 export { createTaskResumeInfoHook } from "./task-resume-info";
 export { createStartWorkHook } from "./start-work";
-export { createSisyphusOrchestratorHook } from "./sisyphus-orchestrator";
+export { createAtlasHook } from "./atlas";
 export { createDelegateTaskRetryHook } from "./delegate-task-retry";

src/hooks/session-recovery/index.test.ts

@@ -199,5 +199,25 @@ describe("detectErrorType", () => {
       // #then should return thinking_block_order
       expect(result).toBe("thinking_block_order")
     })
+
+    it("should detect thinking_block_order even when error message contains tool_use/tool_result in docs URL", () => {
+      // #given Anthropic's extended thinking error with tool_use/tool_result in the documentation text
+      const error = {
+        error: {
+          type: "invalid_request_error",
+          message:
+            "messages.1.content.0.type: Expected `thinking` or `redacted_thinking`, but found `text`. " +
+            "When `thinking` is enabled, a final `assistant` message must start with a thinking block " +
+            "(preceeding the lastmost set of `tool_use` and `tool_result` blocks). " +
+            "We recommend you include thinking blocks from previous turns.",
+        },
+      }
+
+      // #when detectErrorType is called
+      const result = detectErrorType(error)
+
+      // #then should return thinking_block_order (NOT tool_result_missing)
+      expect(result).toBe("thinking_block_order")
+    })
   })
 })

src/hooks/session-recovery/index.ts

@@ -125,10 +125,9 @@ function extractMessageIndex(error: unknown): number | null {
 export function detectErrorType(error: unknown): RecoveryErrorType {
   const message = getErrorMessage(error)
 
-  if (message.includes("tool_use") && message.includes("tool_result")) {
+  // IMPORTANT: Check thinking_block_order BEFORE tool_result_missing
-    return "tool_result_missing"
+  // because Anthropic's extended thinking error messages contain "tool_use" and "tool_result"
-  }
+  // in the documentation URL, which would incorrectly match tool_result_missing
-
   if (
     message.includes("thinking") &&
     (message.includes("first block") ||
@@ -145,6 +144,10 @@ export function detectErrorType(error: unknown): RecoveryErrorType {
     return "thinking_disabled_violation"
   }
 
+  if (message.includes("tool_use") && message.includes("tool_result")) {
+    return "tool_result_missing"
+  }
+
   return null
 }
 

src/hooks/start-work/index.test.ts

@@ -379,24 +379,24 @@ describe("start-work hook", () => {
   })
 
   describe("session agent management", () => {
-    test("should clear session agent when start-work command is triggered", async () => {
+    test("should update session agent to Atlas when start-work command is triggered", async () => {
-      // #given - spy on clearSessionAgent
+      // #given
-      const clearSpy = spyOn(sessionState, "clearSessionAgent")
+      const updateSpy = spyOn(sessionState, "updateSessionAgent")
       
       const hook = createStartWorkHook(createMockPluginInput())
       const output = {
         parts: [{ type: "text", text: "<session-context></session-context>" }],
       }
 
-      // #when - start-work command is processed
+      // #when
       await hook["chat.message"](
         { sessionID: "ses-prometheus-to-sisyphus" },
         output
       )
 
-      // #then - clearSessionAgent should be called with the sessionID
+      // #then
-      expect(clearSpy).toHaveBeenCalledWith("ses-prometheus-to-sisyphus")
+      expect(updateSpy).toHaveBeenCalledWith("ses-prometheus-to-sisyphus", "Atlas")
-      clearSpy.mockRestore()
+      updateSpy.mockRestore()
     })
   })
 })

src/hooks/start-work/index.ts

@@ -10,7 +10,7 @@ import {
   clearBoulderState,
 } from "../../features/boulder-state"
 import { log } from "../../shared/logger"
-import { clearSessionAgent } from "../../features/claude-code-session-state"
+import { updateSessionAgent } from "../../features/claude-code-session-state"
 
 export const HOOK_NAME = "start-work"
 
@@ -71,8 +71,7 @@ export function createStartWorkHook(ctx: PluginInput) {
         sessionID: input.sessionID,
       })
 
-      // Clear previous session agent (e.g., Prometheus) to allow mode transition
+      updateSessionAgent(input.sessionID, "Atlas")
-      clearSessionAgent(input.sessionID)
 
       const existingState = readBoulderState(ctx.directory)
       const sessionId = input.sessionID

src/index.ts

@@ -29,7 +29,7 @@ import {
   createDelegateTaskRetryHook,
   createTaskResumeInfoHook,
   createStartWorkHook,
-  createSisyphusOrchestratorHook,
+  createAtlasHook,
   createPrometheusMdOnlyHook,
 } from "./hooks";
 import {
@@ -198,8 +198,8 @@ const OhMyOpenCodePlugin: Plugin = async (ctx) => {
     ? createStartWorkHook(ctx)
     : null;
 
-  const sisyphusOrchestrator = isHookEnabled("sisyphus-orchestrator")
+  const atlasHook = isHookEnabled("atlas")
-    ? createSisyphusOrchestratorHook(ctx)
+    ? createAtlasHook(ctx)
     : null;
 
   const prometheusMdOnly = isHookEnabled("prometheus-md-only")
@@ -411,7 +411,7 @@ const OhMyOpenCodePlugin: Plugin = async (ctx) => {
       await agentUsageReminder?.event(input);
       await interactiveBashSession?.event(input);
       await ralphLoop?.event(input);
-      await sisyphusOrchestrator?.handler(input);
+      await atlasHook?.handler(input);
 
       const { event } = input;
       const props = event.properties as Record<string, unknown> | undefined;
@@ -565,7 +565,7 @@ const OhMyOpenCodePlugin: Plugin = async (ctx) => {
       await interactiveBashSession?.["tool.execute.after"](input, output);
 await editErrorRecovery?.["tool.execute.after"](input, output);
         await delegateTaskRetry?.["tool.execute.after"](input, output);
-        await sisyphusOrchestrator?.["tool.execute.after"]?.(input, output);
+        await atlasHook?.["tool.execute.after"]?.(input, output);
       await taskResumeInfo["tool.execute.after"](input, output);
     },
   };

src/plugin-handlers/config-handler.test.ts

@@ -10,10 +10,10 @@ describe("Prometheus category config resolution", () => {
     // #when
     const config = resolveCategoryConfig(categoryName)
 
-    // #then - DEFAULT_CATEGORIES only has temperature, not model
+    // #then
     expect(config).toBeDefined()
-    expect(config?.model).toBeUndefined()
+    expect(config?.model).toBe("openai/gpt-5.2-codex")
-    expect(config?.temperature).toBe(0.1)
+    expect(config?.variant).toBe("xhigh")
   })
 
   test("resolves visual-engineering category config", () => {
@@ -23,10 +23,9 @@ describe("Prometheus category config resolution", () => {
     // #when
     const config = resolveCategoryConfig(categoryName)
 
-    // #then - DEFAULT_CATEGORIES only has temperature, not model
+    // #then
     expect(config).toBeDefined()
-    expect(config?.model).toBeUndefined()
+    expect(config?.model).toBe("google/gemini-3-pro-preview")
-    expect(config?.temperature).toBe(0.7)
   })
 
   test("user categories override default categories", () => {
@@ -71,10 +70,10 @@ describe("Prometheus category config resolution", () => {
     // #when
     const config = resolveCategoryConfig(categoryName, userCategories)
 
-    // #then - falls back to DEFAULT_CATEGORIES which has no model
+    // #then - falls back to DEFAULT_CATEGORIES
     expect(config).toBeDefined()
-    expect(config?.model).toBeUndefined()
+    expect(config?.model).toBe("openai/gpt-5.2-codex")
-    expect(config?.temperature).toBe(0.1)
+    expect(config?.variant).toBe("xhigh")
   })
 
   test("preserves all category properties (temperature, top_p, tools, etc.)", () => {

src/plugin-handlers/config-handler.ts

@@ -24,6 +24,7 @@ import type { OhMyOpenCodeConfig } from "../config";
 import { log } from "../shared";
 import { getOpenCodeConfigPaths } from "../shared/opencode-config-dir";
 import { migrateAgentConfig } from "../shared/permission-compat";
+import { AGENT_NAME_MAP } from "../shared/migration";
 import { PROMETHEUS_SYSTEM_PROMPT, PROMETHEUS_PERMISSION } from "../agents/prometheus-prompt";
 import { DEFAULT_CATEGORIES } from "../tools/delegate-task/constants";
 import type { ModelCacheState } from "../plugin-state";
@@ -110,8 +111,13 @@ export function createConfigHandler(deps: ConfigHandlerDeps) {
       )
     }
 
+    // Migrate disabled_agents from old names to new names
+    const migratedDisabledAgents = (pluginConfig.disabled_agents ?? []).map(agent => {
+      return AGENT_NAME_MAP[agent.toLowerCase()] ?? AGENT_NAME_MAP[agent] ?? agent
+    }) as typeof pluginConfig.disabled_agents
+
     const builtinAgents = createBuiltinAgents(
-      pluginConfig.disabled_agents,
+      migratedDisabledAgents,
       pluginConfig.agents,
       ctx.directory,
       config.model as string | undefined,
@@ -154,7 +160,7 @@ export function createConfigHandler(deps: ConfigHandlerDeps) {
       explore?: { tools?: Record<string, unknown> };
       librarian?: { tools?: Record<string, unknown> };
       "multimodal-looker"?: { tools?: Record<string, unknown> };
-      "orchestrator-sisyphus"?: { tools?: Record<string, unknown> };
+      Atlas?: { tools?: Record<string, unknown> };
       Sisyphus?: { tools?: Record<string, unknown> };
     };
     const configAgent = config.agent as AgentConfig | undefined;
@@ -254,6 +260,10 @@ export function createConfigHandler(deps: ConfigHandlerDeps) {
             .filter(([key]) => {
               if (key === "build") return false;
               if (key === "plan" && replacePlan) return false;
+              // Filter out agents that oh-my-opencode provides to prevent
+              // OpenCode defaults from overwriting user config in oh-my-opencode.json
+              // See: https://github.com/code-yeongyu/oh-my-opencode/issues/472
+              if (key in builtinAgents) return false;
               return true;
             })
             .map(([key, value]) => [
@@ -313,17 +323,17 @@ export function createConfigHandler(deps: ConfigHandlerDeps) {
       const agent = agentResult["multimodal-looker"] as AgentWithPermission;
       agent.permission = { ...agent.permission, task: "deny", look_at: "deny" };
     }
-    if (agentResult["orchestrator-sisyphus"]) {
+    if (agentResult["Atlas"]) {
-      const agent = agentResult["orchestrator-sisyphus"] as AgentWithPermission;
+      const agent = agentResult["Atlas"] as AgentWithPermission;
       agent.permission = { ...agent.permission, task: "deny", call_omo_agent: "deny", delegate_task: "allow" };
     }
     if (agentResult.Sisyphus) {
       const agent = agentResult.Sisyphus as AgentWithPermission;
-      agent.permission = { ...agent.permission, call_omo_agent: "deny", delegate_task: "allow" };
+      agent.permission = { ...agent.permission, call_omo_agent: "deny", delegate_task: "allow", question: "allow" };
     }
     if (agentResult["Prometheus (Planner)"]) {
       const agent = agentResult["Prometheus (Planner)"] as AgentWithPermission;
-      agent.permission = { ...agent.permission, call_omo_agent: "deny", delegate_task: "allow" };
+      agent.permission = { ...agent.permission, call_omo_agent: "deny", delegate_task: "allow", question: "allow" };
     }
     if (agentResult["Sisyphus-Junior"]) {
       const agent = agentResult["Sisyphus-Junior"] as AgentWithPermission;

src/shared/agent-tool-restrictions.ts

@@ -28,18 +28,6 @@ const AGENT_RESTRICTIONS: Record<string, Record<string, boolean>> = {
     read: true,
   },
 
-  "document-writer": {
-    task: false,
-    delegate_task: false,
-    call_omo_agent: false,
-  },
-
-  "frontend-ui-ux-engineer": {
-    task: false,
-    delegate_task: false,
-    call_omo_agent: false,
-  },
-
   "Sisyphus-Junior": {
     task: false,
     delegate_task: false,

src/shared/migration.test.ts

@@ -64,7 +64,7 @@ describe("migrateAgentNames", () => {
     // #then: Case-insensitive lookup should migrate correctly
     expect(migrated["Sisyphus"]).toEqual({ model: "test" })
     expect(migrated["Prometheus (Planner)"]).toEqual({ prompt: "test" })
-    expect(migrated["orchestrator-sisyphus"]).toEqual({ model: "openai/gpt-5.2" })
+    expect(migrated["Atlas"]).toEqual({ model: "openai/gpt-5.2" })
   })
 
   test("passes through unknown agent names unchanged", () => {
@@ -80,6 +80,36 @@ describe("migrateAgentNames", () => {
     expect(changed).toBe(false)
     expect(migrated["custom-agent"]).toEqual({ model: "custom/model" })
   })
+
+  test("migrates orchestrator-sisyphus to Atlas", () => {
+    // #given: Config with legacy orchestrator-sisyphus agent name
+    const agents = {
+      "orchestrator-sisyphus": { model: "anthropic/claude-opus-4-5" },
+    }
+
+    // #when: Migrate agent names
+    const { migrated, changed } = migrateAgentNames(agents)
+
+    // #then: orchestrator-sisyphus should be migrated to Atlas
+    expect(changed).toBe(true)
+    expect(migrated["Atlas"]).toEqual({ model: "anthropic/claude-opus-4-5" })
+    expect(migrated["orchestrator-sisyphus"]).toBeUndefined()
+  })
+
+  test("migrates lowercase atlas to Atlas", () => {
+    // #given: Config with lowercase atlas agent name
+    const agents = {
+      atlas: { model: "anthropic/claude-opus-4-5" },
+    }
+
+    // #when: Migrate agent names
+    const { migrated, changed } = migrateAgentNames(agents)
+
+    // #then: lowercase atlas should be migrated to Atlas
+    expect(changed).toBe(true)
+    expect(migrated["Atlas"]).toEqual({ model: "anthropic/claude-opus-4-5" })
+    expect(migrated["atlas"]).toBeUndefined()
+  })
 })
 
 describe("migrateHookNames", () => {
@@ -310,7 +340,7 @@ describe("migrateAgentConfigToCategory", () => {
       { model: "anthropic/claude-sonnet-4-5" },
     ]
 
-    const expectedCategories = ["visual-engineering", "ultrabrain", "quick", "most-capable", "general"]
+    const expectedCategories = ["visual-engineering", "ultrabrain", "quick", "unspecified-high", "unspecified-low"]
 
     // #when: Migrate each config
     const results = configs.map(migrateAgentConfigToCategory)
@@ -370,10 +400,9 @@ describe("shouldDeleteAgentConfig", () => {
 
   test("returns true when all fields match category defaults", () => {
     // #given: Config with fields matching category defaults
-    // Note: DEFAULT_CATEGORIES only has temperature, not model
     const config = {
       category: "visual-engineering",
-      temperature: 0.7,
+      model: "google/gemini-3-pro-preview",
     }
 
     // #when: Check if config should be deleted
@@ -384,10 +413,10 @@ describe("shouldDeleteAgentConfig", () => {
   })
 
   test("returns false when fields differ from category defaults", () => {
-    // #given: Config with custom temperature override
+    // #given: Config with custom model override
     const config = {
       category: "visual-engineering",
-      temperature: 0.9, // Different from default (0.7)
+      model: "anthropic/claude-opus-4-5",
     }
 
     // #when: Check if config should be deleted
@@ -400,10 +429,10 @@ describe("shouldDeleteAgentConfig", () => {
   test("handles different categories with their defaults", () => {
     // #given: Configs for different categories
     const configs = [
-      { category: "ultrabrain", temperature: 0.1 },
+      { category: "ultrabrain" },
-      { category: "quick", temperature: 0.3 },
+      { category: "quick" },
-      { category: "most-capable", temperature: 0.1 },
+      { category: "unspecified-high" },
-      { category: "general", temperature: 0.3 },
+      { category: "unspecified-low" },
     ]
 
     // #when: Check each config

src/shared/migration.ts

@@ -17,10 +17,9 @@ export const AGENT_NAME_MAP: Record<string, string> = {
   oracle: "oracle",
   librarian: "librarian",
   explore: "explore",
-  "frontend-ui-ux-engineer": "frontend-ui-ux-engineer",
-  "document-writer": "document-writer",
   "multimodal-looker": "multimodal-looker",
-  "orchestrator-sisyphus": "orchestrator-sisyphus",
+  "orchestrator-sisyphus": "Atlas",
+  atlas: "Atlas",
 }
 
 export const BUILTIN_AGENT_NAMES = new Set([
@@ -28,13 +27,11 @@ export const BUILTIN_AGENT_NAMES = new Set([
   "oracle",
   "librarian",
   "explore",
-  "frontend-ui-ux-engineer",
-  "document-writer",
   "multimodal-looker",
   "Metis (Plan Consultant)",
   "Momus (Plan Reviewer)",
   "Prometheus (Planner)",
-  "orchestrator-sisyphus",
+  "Atlas",
   "build",
 ])
 
@@ -52,7 +49,7 @@ export const HOOK_NAME_MAP: Record<string, string> = {
  * from explicit model configs to category-based configs.
  * 
  * DO NOT add new entries here. New agents should use:
- * - Category-based config (preferred): { category: "most-capable" }
+ * - Category-based config (preferred): { category: "unspecified-high" }
  * - Or inherit from OpenCode's config.model
  * 
  * This map will be removed in a future major version once migration period ends.
@@ -61,8 +58,8 @@ export const MODEL_TO_CATEGORY_MAP: Record<string, string> = {
   "google/gemini-3-pro-preview": "visual-engineering",
   "openai/gpt-5.2": "ultrabrain",
   "anthropic/claude-haiku-4-5": "quick",
-  "anthropic/claude-opus-4-5": "most-capable",
+  "anthropic/claude-opus-4-5": "unspecified-high",
-  "anthropic/claude-sonnet-4-5": "general",
+  "anthropic/claude-sonnet-4-5": "unspecified-low",
 }
 
 export function migrateAgentNames(agents: Record<string, unknown>): { migrated: Record<string, unknown>; changed: boolean } {
@@ -153,6 +150,22 @@ export function migrateConfigFile(configPath: string, rawConfig: Record<string,
     needsWrite = true
   }
 
+  if (rawConfig.disabled_agents && Array.isArray(rawConfig.disabled_agents)) {
+    const migrated: string[] = []
+    let changed = false
+    for (const agent of rawConfig.disabled_agents as string[]) {
+      const newAgent = AGENT_NAME_MAP[agent.toLowerCase()] ?? AGENT_NAME_MAP[agent] ?? agent
+      if (newAgent !== agent) {
+        changed = true
+      }
+      migrated.push(newAgent)
+    }
+    if (changed) {
+      rawConfig.disabled_agents = migrated
+      needsWrite = true
+    }
+  }
+
   if (rawConfig.disabled_hooks && Array.isArray(rawConfig.disabled_hooks)) {
     const { migrated, changed } = migrateHookNames(rawConfig.disabled_hooks as string[])
     if (changed) {

src/tools/delegate-task/constants.ts

@@ -99,20 +99,42 @@ EXPECTED OUTPUT:
 If your prompt lacks this structure, REWRITE IT before delegating.
 </Caller_Warning>`
 
-export const MOST_CAPABLE_CATEGORY_PROMPT_APPEND = `<Category_Context>
+export const UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND = `<Category_Context>
-You are working on COMPLEX / MOST-CAPABLE tasks.
+You are working on tasks that don't fit specific categories but require moderate effort.
 
-Maximum capability mindset:
+<Selection_Gate>
-- Bring full reasoning power to bear
+BEFORE selecting this category, VERIFY ALL conditions:
-- Consider all edge cases and implications
+1. Task does NOT fit: quick (trivial), visual-engineering (UI), ultrabrain (deep logic), artistry (creative), writing (docs)
-- Deep analysis before action
+2. Task requires more than trivial effort but is NOT system-wide
-- Quality over speed
+3. Scope is contained within a few files/modules
 
-Approach:
+If task fits ANY other category, DO NOT select unspecified-low.
-- Thorough understanding first
+This is NOT a default choice - it's for genuinely unclassifiable moderate-effort work.
-- Comprehensive solution design
+</Selection_Gate>
-- Meticulous execution
+</Category_Context>
-- This is for the most challenging problems
+
+<Caller_Warning>
+THIS CATEGORY USES A MID-TIER MODEL (claude-sonnet-4-5).
+
+**PROVIDE CLEAR STRUCTURE:**
+1. MUST DO: Enumerate required actions explicitly
+2. MUST NOT DO: State forbidden actions to prevent scope creep
+3. EXPECTED OUTPUT: Define concrete success criteria
+</Caller_Warning>`
+
+export const UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND = `<Category_Context>
+You are working on tasks that don't fit specific categories but require substantial effort.
+
+<Selection_Gate>
+BEFORE selecting this category, VERIFY ALL conditions:
+1. Task does NOT fit: quick (trivial), visual-engineering (UI), ultrabrain (deep logic), artistry (creative), writing (docs)
+2. Task requires substantial effort across multiple systems/modules
+3. Changes have broad impact or require careful coordination
+4. NOT just "complex" - must be genuinely unclassifiable AND high-effort
+
+If task fits ANY other category, DO NOT select unspecified-high.
+If task is unclassifiable but moderate-effort, use unspecified-low instead.
+</Selection_Gate>
 </Category_Context>`
 
 export const WRITING_CATEGORY_PROMPT_APPEND = `<Category_Context>
@@ -131,80 +153,16 @@ Approach:
 - Documentation, READMEs, articles, technical writing
 </Category_Context>`
 
-export const GENERAL_CATEGORY_PROMPT_APPEND = `<Category_Context>
-You are working on GENERAL tasks.
 
-Balanced execution mindset:
-- Practical, straightforward approach
-- Good enough is good enough
-- Focus on getting things done
-
-Approach:
-- Standard best practices
-- Reasonable trade-offs
-- Efficient completion
-</Category_Context>
-
-<Caller_Warning>
-THIS CATEGORY USES A MID-TIER MODEL (claude-sonnet-4-5).
-
-While capable, this model benefits significantly from EXPLICIT instructions.
-
-**PROVIDE CLEAR STRUCTURE:**
-1. MUST DO: Enumerate required actions explicitly - don't assume inference
-2. MUST NOT DO: State forbidden actions to prevent scope creep or wrong approaches
-3. EXPECTED OUTPUT: Define concrete success criteria and deliverables
-
-**COMMON PITFALLS WITHOUT EXPLICIT INSTRUCTIONS:**
-- Model may take shortcuts that miss edge cases
-- Implicit requirements get overlooked
-- Output format may not match expectations
-- Scope may expand beyond intended boundaries
-
-**RECOMMENDED PROMPT PATTERN:**
-\`\`\`
-TASK: [Clear, single-purpose goal]
-
-CONTEXT: [Relevant background the model needs]
-
-MUST DO:
-- [Explicit requirement 1]
-- [Explicit requirement 2]
-
-MUST NOT DO:
-- [Boundary/constraint 1]
-- [Boundary/constraint 2]
-
-EXPECTED OUTPUT:
-- [What success looks like]
-- [How to verify completion]
-\`\`\`
-
-The more explicit your prompt, the better the results.
-</Caller_Warning>`
 
 export const DEFAULT_CATEGORIES: Record<string, CategoryConfig> = {
-  "visual-engineering": {
+  "visual-engineering": { model: "google/gemini-3-pro-preview" },
-    temperature: 0.7,
+  ultrabrain: { model: "openai/gpt-5.2-codex", variant: "xhigh" },
-  },
+  artistry: { model: "google/gemini-3-pro-preview", variant: "max" },
-  ultrabrain: {
+  quick: { model: "anthropic/claude-haiku-4-5" },
-    temperature: 0.1,
+  "unspecified-low": { model: "anthropic/claude-sonnet-4-5" },
-  },
+  "unspecified-high": { model: "anthropic/claude-opus-4-5", variant: "max" },
-  artistry: {
+  writing: { model: "google/gemini-3-flash-preview" },
-    temperature: 0.9,
-  },
-  quick: {
-    temperature: 0.3,
-  },
-  "most-capable": {
-    temperature: 0.1,
-  },
-  writing: {
-    temperature: 0.5,
-  },
-  general: {
-    temperature: 0.3,
-  },
 }
 
 export const CATEGORY_PROMPT_APPENDS: Record<string, string> = {
@@ -212,19 +170,19 @@ export const CATEGORY_PROMPT_APPENDS: Record<string, string> = {
   ultrabrain: STRATEGIC_CATEGORY_PROMPT_APPEND,
   artistry: ARTISTRY_CATEGORY_PROMPT_APPEND,
   quick: QUICK_CATEGORY_PROMPT_APPEND,
-  "most-capable": MOST_CAPABLE_CATEGORY_PROMPT_APPEND,
+  "unspecified-low": UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND,
+  "unspecified-high": UNSPECIFIED_HIGH_CATEGORY_PROMPT_APPEND,
   writing: WRITING_CATEGORY_PROMPT_APPEND,
-  general: GENERAL_CATEGORY_PROMPT_APPEND,
 }
 
 export const CATEGORY_DESCRIPTIONS: Record<string, string> = {
   "visual-engineering": "Frontend, UI/UX, design, styling, animation",
-  ultrabrain: "Strict architecture design, very complex business logic",
+  ultrabrain: "Deep logical reasoning, complex architecture decisions requiring extensive analysis",
   artistry: "Highly creative/artistic tasks, novel ideas",
-  quick: "Cheap & fast - small tasks with minimal overhead, budget-friendly",
+  quick: "Trivial tasks - single file changes, typo fixes, simple modifications",
-  "most-capable": "Complex tasks requiring maximum capability",
+  "unspecified-low": "Tasks that don't fit other categories, low effort required",
+  "unspecified-high": "Tasks that don't fit other categories, high effort required",
   writing: "Documentation, prose, technical writing",
-  general: "General purpose tasks",
 }
 
 const BUILTIN_CATEGORIES = Object.keys(DEFAULT_CATEGORIES).join(", ")

src/tools/delegate-task/tools.test.ts

@@ -8,24 +8,23 @@ const SYSTEM_DEFAULT_MODEL = "anthropic/claude-sonnet-4-5"
 
 describe("sisyphus-task", () => {
   describe("DEFAULT_CATEGORIES", () => {
-    test("visual-engineering category has temperature config only (model removed)", () => {
+    test("visual-engineering category has model config", () => {
       // #given
       const category = DEFAULT_CATEGORIES["visual-engineering"]
 
       // #when / #then
       expect(category).toBeDefined()
-      expect(category.model).toBeUndefined()
+      expect(category.model).toBe("google/gemini-3-pro-preview")
-      expect(category.temperature).toBe(0.7)
     })
 
-    test("ultrabrain category has temperature config only (model removed)", () => {
+    test("ultrabrain category has model and variant config", () => {
       // #given
       const category = DEFAULT_CATEGORIES["ultrabrain"]
 
       // #when / #then
       expect(category).toBeDefined()
-      expect(category.model).toBeUndefined()
+      expect(category.model).toBe("openai/gpt-5.2-codex")
-      expect(category.temperature).toBe(0.1)
+      expect(category.variant).toBe("xhigh")
     })
   })
 
@@ -61,13 +60,13 @@ describe("sisyphus-task", () => {
       }
     })
 
-    test("most-capable category exists and has description", () => {
+    test("unspecified-high category exists and has description", () => {
       // #given / #when
-      const description = CATEGORY_DESCRIPTIONS["most-capable"]
+      const description = CATEGORY_DESCRIPTIONS["unspecified-high"]
 
       // #then
       expect(description).toBeDefined()
-      expect(description).toContain("Complex")
+      expect(description).toContain("high effort")
     })
   })
 
@@ -141,16 +140,16 @@ describe("sisyphus-task", () => {
       expect(result).toBeNull()
     })
 
-    test("returns systemDefaultModel for builtin category (categories no longer have default models)", () => {
+    test("returns default model from DEFAULT_CATEGORIES for builtin category", () => {
       // #given
       const categoryName = "visual-engineering"
 
       // #when
       const result = resolveCategoryConfig(categoryName, { systemDefaultModel: SYSTEM_DEFAULT_MODEL })
 
-      // #then - model comes from systemDefaultModel since categories no longer have model defaults
+      // #then
       expect(result).not.toBeNull()
-      expect(result!.config.model).toBe(SYSTEM_DEFAULT_MODEL)
+      expect(result!.config.model).toBe("google/gemini-3-pro-preview")
       expect(result!.promptAppend).toContain("VISUAL/UI")
     })
 
@@ -227,21 +226,21 @@ describe("sisyphus-task", () => {
       expect(result!.config.temperature).toBe(0.3)
     })
 
-    test("inheritedModel takes precedence over systemDefaultModel", () => {
+    test("category built-in model takes precedence over inheritedModel", () => {
-      // #given - builtin category, parent model provided
+      // #given - builtin category with its own model, parent model also provided
       const categoryName = "visual-engineering"
       const inheritedModel = "cliproxy/claude-opus-4-5"
 
       // #when
       const result = resolveCategoryConfig(categoryName, { inheritedModel, systemDefaultModel: SYSTEM_DEFAULT_MODEL })
 
-      // #then - inheritedModel wins over systemDefaultModel
+      // #then - category's built-in model wins over inheritedModel
       expect(result).not.toBeNull()
-      expect(result!.config.model).toBe("cliproxy/claude-opus-4-5")
+      expect(result!.config.model).toBe("google/gemini-3-pro-preview")
     })
 
-    test("inheritedModel is used as fallback when category has no user model", () => {
+    test("systemDefaultModel is used as fallback when custom category has no model", () => {
-      // #given - custom category with no model defined, only inheritedModel as fallback
+      // #given - custom category with no model defined
       const categoryName = "my-custom-no-model"
       const userCategories = { "my-custom-no-model": { temperature: 0.5 } } as unknown as Record<string, CategoryConfig>
       const inheritedModel = "cliproxy/claude-opus-4-5"
@@ -249,9 +248,9 @@ describe("sisyphus-task", () => {
       // #when
       const result = resolveCategoryConfig(categoryName, { userCategories, inheritedModel, systemDefaultModel: SYSTEM_DEFAULT_MODEL })
 
-      // #then - parent model is used as fallback since custom category has no user model
+      // #then - systemDefaultModel is used since custom category has no built-in model
       expect(result).not.toBeNull()
-      expect(result!.config.model).toBe("cliproxy/claude-opus-4-5")
+      expect(result!.config.model).toBe(SYSTEM_DEFAULT_MODEL)
     })
 
     test("user model takes precedence over inheritedModel", () => {
@@ -270,7 +269,7 @@ describe("sisyphus-task", () => {
       expect(result!.config.model).toBe("my-provider/my-model")
     })
 
-    test("systemDefaultModel is used when no user model and no inheritedModel", () => {
+    test("default model from category config is used when no user model and no inheritedModel", () => {
       // #given
       const categoryName = "visual-engineering"
 
@@ -279,7 +278,7 @@ describe("sisyphus-task", () => {
 
       // #then
       expect(result).not.toBeNull()
-      expect(result!.config.model).toBe(SYSTEM_DEFAULT_MODEL)
+      expect(result!.config.model).toBe("google/gemini-3-pro-preview")
     })
   })
 
@@ -346,6 +345,124 @@ describe("sisyphus-task", () => {
         variant: "xhigh",
       })
     })
+
+    test("DEFAULT_CATEGORIES variant passes to background WITHOUT userCategories", async () => {
+      // #given - NO userCategories, testing DEFAULT_CATEGORIES only
+      const { createDelegateTask } = require("./tools")
+      let launchInput: any
+
+      const mockManager = {
+        launch: async (input: any) => {
+          launchInput = input
+          return {
+            id: "task-default-variant",
+            sessionID: "session-default-variant",
+            description: "Default variant task",
+            agent: "Sisyphus-Junior",
+            status: "running",
+          }
+        },
+      }
+
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          create: async () => ({ data: { id: "test-session" } }),
+          prompt: async () => ({ data: {} }),
+          messages: async () => ({ data: [] }),
+        },
+      }
+
+      // NO userCategories - must use DEFAULT_CATEGORIES
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+      })
+
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+
+      // #when - unspecified-high has variant: "max" in DEFAULT_CATEGORIES
+      await tool.execute(
+        {
+          description: "Test unspecified-high default variant",
+          prompt: "Do something",
+          category: "unspecified-high",
+          run_in_background: true,
+          skills: [],
+        },
+        toolContext
+      )
+
+      // #then - variant MUST be "max" from DEFAULT_CATEGORIES
+      expect(launchInput.model).toEqual({
+        providerID: "anthropic",
+        modelID: "claude-opus-4-5",
+        variant: "max",
+      })
+    })
+
+    test("DEFAULT_CATEGORIES variant passes to sync session.prompt WITHOUT userCategories", async () => {
+      // #given - NO userCategories, testing DEFAULT_CATEGORIES for sync mode
+      const { createDelegateTask } = require("./tools")
+      let promptBody: any
+
+      const mockManager = { launch: async () => ({}) }
+
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          get: async () => ({ data: { directory: "/project" } }),
+          create: async () => ({ data: { id: "ses_sync_default_variant" } }),
+          prompt: async (input: any) => {
+            promptBody = input.body
+            return { data: {} }
+          },
+          messages: async () => ({
+            data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: "done" }] }]
+          }),
+          status: async () => ({ data: { "ses_sync_default_variant": { type: "idle" } } }),
+        },
+      }
+
+      // NO userCategories - must use DEFAULT_CATEGORIES
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+      })
+
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+
+      // #when - unspecified-high has variant: "max" in DEFAULT_CATEGORIES
+      await tool.execute(
+        {
+          description: "Test unspecified-high sync variant",
+          prompt: "Do something",
+          category: "unspecified-high",
+          run_in_background: false,
+          skills: [],
+        },
+        toolContext
+      )
+
+      // #then - variant MUST be "max" from DEFAULT_CATEGORIES
+      expect(promptBody.model).toEqual({
+        providerID: "anthropic",
+        modelID: "claude-opus-4-5",
+        variant: "max",
+      })
+    }, { timeout: 20000 })
   })
 
   describe("skills parameter", () => {
@@ -841,6 +958,389 @@ describe("sisyphus-task", () => {
     }, { timeout: 20000 })
   })
 
+  describe("unstable agent forced background mode", () => {
+    test("gemini model with run_in_background=false should force background but wait for result", async () => {
+      // #given - category using gemini model with run_in_background=false
+      const { createDelegateTask } = require("./tools")
+      let launchCalled = false
+      
+      const mockManager = {
+        launch: async () => {
+          launchCalled = true
+          return {
+            id: "task-unstable",
+            sessionID: "ses_unstable_gemini",
+            description: "Unstable gemini task",
+            agent: "Sisyphus-Junior",
+            status: "running",
+          }
+        },
+      }
+      
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          get: async () => ({ data: { directory: "/project" } }),
+          create: async () => ({ data: { id: "ses_unstable_gemini" } }),
+          prompt: async () => ({ data: {} }),
+          messages: async () => ({
+            data: [
+              { info: { role: "assistant", time: { created: Date.now() } }, parts: [{ type: "text", text: "Gemini task completed successfully" }] }
+            ]
+          }),
+          status: async () => ({ data: { "ses_unstable_gemini": { type: "idle" } } }),
+        },
+      }
+      
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+      })
+      
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+      
+      // #when - using visual-engineering (gemini model) with run_in_background=false
+      const result = await tool.execute(
+        {
+          description: "Test gemini forced background",
+          prompt: "Do something visual",
+          category: "visual-engineering",
+          run_in_background: false,
+          skills: [],
+        },
+        toolContext
+      )
+      
+      // #then - should launch as background BUT wait for and return actual result
+      expect(launchCalled).toBe(true)
+      expect(result).toContain("UNSTABLE AGENT")
+      expect(result).toContain("Gemini task completed successfully")
+    }, { timeout: 20000 })
+
+    test("gemini model with run_in_background=true should not show unstable message (normal background)", async () => {
+      // #given - category using gemini model with run_in_background=true (normal background flow)
+      const { createDelegateTask } = require("./tools")
+      let launchCalled = false
+      
+      const mockManager = {
+        launch: async () => {
+          launchCalled = true
+          return {
+            id: "task-normal-bg",
+            sessionID: "ses_normal_bg",
+            description: "Normal background task",
+            agent: "Sisyphus-Junior",
+            status: "running",
+          }
+        },
+      }
+      
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          create: async () => ({ data: { id: "test-session" } }),
+          prompt: async () => ({ data: {} }),
+          messages: async () => ({ data: [] }),
+        },
+      }
+      
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+      })
+      
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+      
+      // #when - using visual-engineering with run_in_background=true (normal background)
+      const result = await tool.execute(
+        {
+          description: "Test normal background",
+          prompt: "Do something visual",
+          category: "visual-engineering",
+          run_in_background: true,  // User explicitly says true - normal background
+          skills: [],
+        },
+        toolContext
+      )
+      
+      // #then - should NOT show unstable message (it's normal background flow)
+      expect(launchCalled).toBe(true)
+      expect(result).not.toContain("UNSTABLE AGENT MODE")
+      expect(result).toContain("task-normal-bg")
+    })
+
+    test("non-gemini model with run_in_background=false should run sync (not forced to background)", async () => {
+      // #given - category using non-gemini model with run_in_background=false
+      const { createDelegateTask } = require("./tools")
+      let launchCalled = false
+      let promptCalled = false
+      
+      const mockManager = {
+        launch: async () => {
+          launchCalled = true
+          return { id: "should-not-be-called", sessionID: "x", description: "x", agent: "x", status: "running" }
+        },
+      }
+      
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          get: async () => ({ data: { directory: "/project" } }),
+          create: async () => ({ data: { id: "ses_sync_non_gemini" } }),
+          prompt: async () => {
+            promptCalled = true
+            return { data: {} }
+          },
+          messages: async () => ({
+            data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: "Done sync" }] }]
+          }),
+          status: async () => ({ data: { "ses_sync_non_gemini": { type: "idle" } } }),
+        },
+      }
+      
+      // Use ultrabrain which uses gpt-5.2 (non-gemini)
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+      })
+      
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+      
+      // #when - using ultrabrain (gpt model) with run_in_background=false
+      const result = await tool.execute(
+        {
+          description: "Test non-gemini sync",
+          prompt: "Do something smart",
+          category: "ultrabrain",
+          run_in_background: false,
+          skills: [],
+        },
+        toolContext
+      )
+      
+      // #then - should run sync, NOT forced to background
+      expect(launchCalled).toBe(false)  // manager.launch should NOT be called
+      expect(promptCalled).toBe(true)   // sync mode uses session.prompt
+      expect(result).not.toContain("UNSTABLE AGENT MODE")
+    }, { timeout: 20000 })
+
+    test("artistry category (gemini) with run_in_background=false should force background but wait for result", async () => {
+      // #given - artistry also uses gemini model
+      const { createDelegateTask } = require("./tools")
+      let launchCalled = false
+      
+      const mockManager = {
+        launch: async () => {
+          launchCalled = true
+          return {
+            id: "task-artistry",
+            sessionID: "ses_artistry_gemini",
+            description: "Artistry gemini task",
+            agent: "Sisyphus-Junior",
+            status: "running",
+          }
+        },
+      }
+      
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          get: async () => ({ data: { directory: "/project" } }),
+          create: async () => ({ data: { id: "ses_artistry_gemini" } }),
+          prompt: async () => ({ data: {} }),
+          messages: async () => ({
+            data: [
+              { info: { role: "assistant", time: { created: Date.now() } }, parts: [{ type: "text", text: "Artistry result here" }] }
+            ]
+          }),
+          status: async () => ({ data: { "ses_artistry_gemini": { type: "idle" } } }),
+        },
+      }
+      
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+      })
+      
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+      
+      // #when - artistry category (gemini-3-pro-preview with max variant)
+      const result = await tool.execute(
+        {
+          description: "Test artistry forced background",
+          prompt: "Do something artistic",
+          category: "artistry",
+          run_in_background: false,
+          skills: [],
+        },
+        toolContext
+      )
+      
+      // #then - should launch as background BUT wait for and return actual result
+      expect(launchCalled).toBe(true)
+      expect(result).toContain("UNSTABLE AGENT")
+      expect(result).toContain("Artistry result here")
+    }, { timeout: 20000 })
+
+    test("writing category (gemini-flash) with run_in_background=false should force background but wait for result", async () => {
+      // #given - writing uses gemini-3-flash-preview
+      const { createDelegateTask } = require("./tools")
+      let launchCalled = false
+      
+      const mockManager = {
+        launch: async () => {
+          launchCalled = true
+          return {
+            id: "task-writing",
+            sessionID: "ses_writing_gemini",
+            description: "Writing gemini task",
+            agent: "Sisyphus-Junior",
+            status: "running",
+          }
+        },
+      }
+      
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          get: async () => ({ data: { directory: "/project" } }),
+          create: async () => ({ data: { id: "ses_writing_gemini" } }),
+          prompt: async () => ({ data: {} }),
+          messages: async () => ({
+            data: [
+              { info: { role: "assistant", time: { created: Date.now() } }, parts: [{ type: "text", text: "Writing result here" }] }
+            ]
+          }),
+          status: async () => ({ data: { "ses_writing_gemini": { type: "idle" } } }),
+        },
+      }
+      
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+      })
+      
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+      
+      // #when - writing category (gemini-3-flash-preview)
+      const result = await tool.execute(
+        {
+          description: "Test writing forced background",
+          prompt: "Write something",
+          category: "writing",
+          run_in_background: false,
+          skills: [],
+        },
+        toolContext
+      )
+      
+      // #then - should launch as background BUT wait for and return actual result
+      expect(launchCalled).toBe(true)
+      expect(result).toContain("UNSTABLE AGENT")
+      expect(result).toContain("Writing result here")
+    }, { timeout: 20000 })
+
+    test("is_unstable_agent=true should force background but wait for result", async () => {
+      // #given - custom category with is_unstable_agent=true but non-gemini model
+      const { createDelegateTask } = require("./tools")
+      let launchCalled = false
+      
+      const mockManager = {
+        launch: async () => {
+          launchCalled = true
+          return {
+            id: "task-custom-unstable",
+            sessionID: "ses_custom_unstable",
+            description: "Custom unstable task",
+            agent: "Sisyphus-Junior",
+            status: "running",
+          }
+        },
+      }
+      
+      const mockClient = {
+        app: { agents: async () => ({ data: [] }) },
+        config: { get: async () => ({ data: { model: SYSTEM_DEFAULT_MODEL } }) },
+        session: {
+          get: async () => ({ data: { directory: "/project" } }),
+          create: async () => ({ data: { id: "ses_custom_unstable" } }),
+          prompt: async () => ({ data: {} }),
+          messages: async () => ({
+            data: [
+              { info: { role: "assistant", time: { created: Date.now() } }, parts: [{ type: "text", text: "Custom unstable result" }] }
+            ]
+          }),
+          status: async () => ({ data: { "ses_custom_unstable": { type: "idle" } } }),
+        },
+      }
+      
+      const tool = createDelegateTask({
+        manager: mockManager,
+        client: mockClient,
+        userCategories: {
+          "my-unstable-cat": {
+            model: "openai/gpt-5.2",
+            is_unstable_agent: true,
+          },
+        },
+      })
+      
+      const toolContext = {
+        sessionID: "parent-session",
+        messageID: "parent-message",
+        agent: "Sisyphus",
+        abort: new AbortController().signal,
+      }
+      
+      // #when - using custom unstable category with run_in_background=false
+      const result = await tool.execute(
+        {
+          description: "Test custom unstable",
+          prompt: "Do something",
+          category: "my-unstable-cat",
+          run_in_background: false,
+          skills: [],
+        },
+        toolContext
+      )
+      
+      // #then - should launch as background BUT wait for and return actual result
+      expect(launchCalled).toBe(true)
+      expect(result).toContain("UNSTABLE AGENT")
+      expect(result).toContain("Custom unstable result")
+    }, { timeout: 20000 })
+  })
+
   describe("buildSystemContent", () => {
     test("returns undefined when no skills and no category promptAppend", () => {
       // #given
@@ -894,31 +1394,43 @@ describe("sisyphus-task", () => {
   })
 
   describe("modelInfo detection via resolveCategoryConfig", () => {
-    test("systemDefaultModel is used when no userModel and no inheritedModel", () => {
+    test("catalog model is used for category with catalog entry", () => {
-      // #given - builtin category, no user model, no inherited model
+      // #given - ultrabrain has catalog entry
       const categoryName = "ultrabrain"
       
       // #when
       const resolved = resolveCategoryConfig(categoryName, { systemDefaultModel: SYSTEM_DEFAULT_MODEL })
       
-      // #then - actualModel should be systemDefaultModel (categories no longer have model defaults)
+      // #then - catalog model is used
       expect(resolved).not.toBeNull()
-      const actualModel = resolved!.config.model
+      expect(resolved!.config.model).toBe("openai/gpt-5.2-codex")
-      expect(actualModel).toBe(SYSTEM_DEFAULT_MODEL)
+      expect(resolved!.config.variant).toBe("xhigh")
     })
 
-    test("inheritedModel takes precedence over systemDefaultModel for builtin category", () => {
+    test("default model is used for category with default entry", () => {
-      // #given - builtin ultrabrain category, inherited model from parent
+      // #given - unspecified-low has default model
+      const categoryName = "unspecified-low"
+      
+      // #when
+      const resolved = resolveCategoryConfig(categoryName, { systemDefaultModel: SYSTEM_DEFAULT_MODEL })
+      
+      // #then - default model from DEFAULT_CATEGORIES is used
+      expect(resolved).not.toBeNull()
+      expect(resolved!.config.model).toBe("anthropic/claude-sonnet-4-5")
+    })
+
+    test("category built-in model takes precedence over inheritedModel for builtin category", () => {
+      // #given - builtin ultrabrain category with its own model, inherited model also provided
       const categoryName = "ultrabrain"
       const inheritedModel = "cliproxy/claude-opus-4-5"
       
       // #when
       const resolved = resolveCategoryConfig(categoryName, { inheritedModel, systemDefaultModel: SYSTEM_DEFAULT_MODEL })
       
-      // #then - inheritedModel wins over systemDefaultModel
+      // #then - category's built-in model wins (ultrabrain uses gpt-5.2-codex)
       expect(resolved).not.toBeNull()
       const actualModel = resolved!.config.model
-      expect(actualModel).toBe("cliproxy/claude-opus-4-5")
+      expect(actualModel).toBe("openai/gpt-5.2-codex")
     })
 
     test("when user defines model - modelInfo should report user-defined regardless of inheritedModel", () => {
@@ -966,18 +1478,18 @@ describe("sisyphus-task", () => {
     // ===== TESTS FOR resolveModel() INTEGRATION (TDD GREEN) =====
     // These tests verify the NEW behavior where categories do NOT have default models
 
-    test("FIXED: inheritedModel takes precedence over systemDefaultModel", () => {
+    test("FIXED: category built-in model takes precedence over inheritedModel", () => {
-      // #given a builtin category, and an inherited model from parent
+      // #given a builtin category with its own model, and an inherited model from parent
-      // The NEW correct chain: userConfig?.model ?? inheritedModel ?? systemDefaultModel
+      // The CORRECT chain: userConfig?.model ?? categoryBuiltIn ?? systemDefaultModel
       const categoryName = "ultrabrain"
-      const inheritedModel = "anthropic/claude-opus-4-5" // inherited from parent session
+      const inheritedModel = "anthropic/claude-opus-4-5"
       
-      // #when userConfig.model is undefined and inheritedModel is set
+      // #when category has a built-in model (gpt-5.2-codex for ultrabrain)
       const resolved = resolveCategoryConfig(categoryName, { inheritedModel, systemDefaultModel: SYSTEM_DEFAULT_MODEL })
       
-      // #then inheritedModel should be used, NOT systemDefaultModel
+      // #then category's built-in model should be used, NOT inheritedModel
       expect(resolved).not.toBeNull()
-      expect(resolved!.model).toBe("anthropic/claude-opus-4-5")
+      expect(resolved!.model).toBe("openai/gpt-5.2-codex")
     })
 
     test("FIXED: systemDefaultModel is used when no userConfig.model and no inheritedModel", () => {
@@ -1016,8 +1528,8 @@ describe("sisyphus-task", () => {
       expect(resolved!.model).toBe("custom/user-model")
     })
 
-    test("FIXED: empty string in userConfig.model is treated as unset and falls back", () => {
+    test("FIXED: empty string in userConfig.model is treated as unset and falls back to systemDefault", () => {
-      // #given userConfig.model is empty string ""
+      // #given userConfig.model is empty string "" for a custom category (no built-in model)
       const categoryName = "custom-empty-model"
       const userCategories = { "custom-empty-model": { model: "", temperature: 0.3 } }
       const inheritedModel = "anthropic/claude-opus-4-5"
@@ -1025,13 +1537,13 @@ describe("sisyphus-task", () => {
       // #when resolveCategoryConfig is called
       const resolved = resolveCategoryConfig(categoryName, { userCategories, inheritedModel, systemDefaultModel: SYSTEM_DEFAULT_MODEL })
       
-      // #then should fall back to inheritedModel since "" is normalized to undefined
+      // #then should fall back to systemDefaultModel since custom category has no built-in model
       expect(resolved).not.toBeNull()
-      expect(resolved!.model).toBe("anthropic/claude-opus-4-5")
+      expect(resolved!.model).toBe(SYSTEM_DEFAULT_MODEL)
     })
 
-    test("FIXED: undefined userConfig.model falls back to inheritedModel", () => {
+    test("FIXED: undefined userConfig.model falls back to category built-in model", () => {
-      // #given user explicitly sets a category but leaves model undefined
+      // #given user sets a builtin category but leaves model undefined
       const categoryName = "visual-engineering"
       // Using type assertion since we're testing fallback behavior for categories without model
       const userCategories = { "visual-engineering": { temperature: 0.2 } } as unknown as Record<string, CategoryConfig>
@@ -1040,9 +1552,9 @@ describe("sisyphus-task", () => {
       // #when resolveCategoryConfig is called
       const resolved = resolveCategoryConfig(categoryName, { userCategories, inheritedModel, systemDefaultModel: SYSTEM_DEFAULT_MODEL })
       
-      // #then should use inheritedModel
+      // #then should use category's built-in model (gemini-3-pro-preview for visual-engineering)
       expect(resolved).not.toBeNull()
-      expect(resolved!.model).toBe("anthropic/claude-opus-4-5")
+      expect(resolved!.model).toBe("google/gemini-3-pro-preview")
     })
 
     test("systemDefaultModel is used when no other model is available", () => {

src/tools/delegate-task/tools.ts

@@ -124,16 +124,18 @@ export function resolveCategoryConfig(
     return null
   }
 
-  // Model priority: user override > inherited from parent > system default
+  // Model priority for categories: user override > category default > system default
+  // Categories have explicit models - no inheritance from parent session
   const model = resolveModel({
     userModel: userConfig?.model,
-    inheritedModel,
+    inheritedModel: defaultConfig?.model, // Category's built-in model takes precedence over system default
     systemDefault: systemDefaultModel,
   })
   const config: CategoryConfig = {
     ...defaultConfig,
     ...userConfig,
     model,
+    variant: userConfig?.variant ?? defaultConfig?.variant,
   }
 
   let promptAppend = defaultPromptAppend
@@ -480,6 +482,121 @@ ${textContent || "(No text output)"}`
             : parsedModel)
           : undefined
         categoryPromptAppend = resolved.promptAppend || undefined
+
+        // Unstable agent detection - launch as background for monitoring but wait for result
+        const isUnstableAgent = resolved.config.is_unstable_agent === true || actualModel.toLowerCase().includes("gemini")
+        if (isUnstableAgent && args.run_in_background === false) {
+          const systemContent = buildSystemContent({ skillContent, categoryPromptAppend })
+
+          try {
+            const task = await manager.launch({
+              description: args.description,
+              prompt: args.prompt,
+              agent: agentToUse,
+              parentSessionID: ctx.sessionID,
+              parentMessageID: ctx.messageID,
+              parentModel,
+              parentAgent,
+              model: categoryModel,
+              skills: args.skills.length > 0 ? args.skills : undefined,
+              skillContent: systemContent,
+            })
+
+            const sessionID = task.sessionID
+            if (!sessionID) {
+              return formatDetailedError(new Error("Background task launched but no sessionID returned"), {
+                operation: "Launch background task (unstable agent)",
+                args,
+                agent: agentToUse,
+                category: args.category,
+              })
+            }
+
+            ctx.metadata?.({
+              title: args.description,
+              metadata: { sessionId: sessionID, category: args.category },
+            })
+
+            const startTime = new Date()
+
+            // Poll for completion (same logic as sync mode)
+            const POLL_INTERVAL_MS = 500
+            const MAX_POLL_TIME_MS = 10 * 60 * 1000
+            const MIN_STABILITY_TIME_MS = 10000
+            const STABILITY_POLLS_REQUIRED = 3
+            const pollStart = Date.now()
+            let lastMsgCount = 0
+            let stablePolls = 0
+
+            while (Date.now() - pollStart < MAX_POLL_TIME_MS) {
+              if (ctx.abort?.aborted) {
+                return `[UNSTABLE AGENT] Task aborted.\n\nSession ID: ${sessionID}`
+              }
+
+              await new Promise(resolve => setTimeout(resolve, POLL_INTERVAL_MS))
+
+              const statusResult = await client.session.status()
+              const allStatuses = (statusResult.data ?? {}) as Record<string, { type: string }>
+              const sessionStatus = allStatuses[sessionID]
+
+              if (sessionStatus && sessionStatus.type !== "idle") {
+                stablePolls = 0
+                lastMsgCount = 0
+                continue
+              }
+
+              if (Date.now() - pollStart < MIN_STABILITY_TIME_MS) continue
+
+              const messagesCheck = await client.session.messages({ path: { id: sessionID } })
+              const msgs = ((messagesCheck as { data?: unknown }).data ?? messagesCheck) as Array<unknown>
+              const currentMsgCount = msgs.length
+
+              if (currentMsgCount === lastMsgCount) {
+                stablePolls++
+                if (stablePolls >= STABILITY_POLLS_REQUIRED) break
+              } else {
+                stablePolls = 0
+                lastMsgCount = currentMsgCount
+              }
+            }
+
+            const messagesResult = await client.session.messages({ path: { id: sessionID } })
+            const messages = ((messagesResult as { data?: unknown }).data ?? messagesResult) as Array<{
+              info?: { role?: string; time?: { created?: number } }
+              parts?: Array<{ type?: string; text?: string }>
+            }>
+
+            const assistantMessages = messages
+              .filter((m) => m.info?.role === "assistant")
+              .sort((a, b) => (b.info?.time?.created ?? 0) - (a.info?.time?.created ?? 0))
+            const lastMessage = assistantMessages[0]
+
+            if (!lastMessage) {
+              return `[UNSTABLE AGENT] No assistant response found.\n\nSession ID: ${sessionID}`
+            }
+
+            const textParts = lastMessage?.parts?.filter((p) => p.type === "text" || p.type === "reasoning") ?? []
+            const textContent = textParts.map((p) => p.text ?? "").filter(Boolean).join("\n")
+            const duration = formatDuration(startTime)
+
+            return `[UNSTABLE AGENT] Task completed in ${duration}.
+
+Model: ${actualModel} (unstable/experimental - launched via background for monitoring)
+Agent: ${agentToUse}${args.category ? ` (category: ${args.category})` : ""}
+Session ID: ${sessionID}
+
+---
+
+${textContent || "(No text output)"}`
+          } catch (error) {
+            return formatDetailedError(error, {
+              operation: "Launch background task (unstable agent)",
+              args,
+              agent: agentToUse,
+              category: args.category,
+            })
+          }
+        }
       } else {
         if (!args.subagent_type?.trim()) {
           return `Agent name cannot be empty.`

src/tools/skill-mcp/tools.ts

@@ -24,7 +24,7 @@ function validateOperationParams(args: SkillMcpArgs): OperationType {
         `Examples:\n` +
         `  skill_mcp(mcp_name="sqlite", tool_name="query", arguments='{"sql": "SELECT * FROM users"}')\n` +
         `  skill_mcp(mcp_name="memory", resource_name="memory://notes")\n` +
-      `  skill_mcp(mcp_name="helper", prompt_name="summarize", arguments='{"text": "..."}')`
+        `  skill_mcp(mcp_name="helper", prompt_name="summarize", arguments='{"text": "..."}')`,
     )
   }
 
@@ -33,12 +33,14 @@ function validateOperationParams(args: SkillMcpArgs): OperationType {
       args.tool_name && `tool_name="${args.tool_name}"`,
       args.resource_name && `resource_name="${args.resource_name}"`,
       args.prompt_name && `prompt_name="${args.prompt_name}"`,
-    ].filter(Boolean).join(", ")
+    ]
+      .filter(Boolean)
+      .join(", ")
 
     throw new Error(
       `Multiple operations specified. Exactly one of tool_name, resource_name, or prompt_name must be provided.\n\n` +
         `Received: ${provided}\n\n` +
-      `Use separate calls for each operation.`
+        `Use separate calls for each operation.`,
     )
   }
 
@@ -47,7 +49,7 @@ function validateOperationParams(args: SkillMcpArgs): OperationType {
 
 function findMcpServer(
   mcpName: string,
-  skills: LoadedSkill[]
+  skills: LoadedSkill[],
 ): { skill: LoadedSkill; config: NonNullable<LoadedSkill["mcpConfig"]>[string] } | null {
   for (const skill of skills) {
     if (skill.mcpConfig && mcpName in skill.mcpConfig) {
@@ -75,7 +77,10 @@ function parseArguments(argsJson: string | Record<string, unknown> | undefined):
     return argsJson
   }
   try {
-    const parsed = JSON.parse(argsJson)
+    // Strip outer single quotes if present (common in LLM output)
+    const jsonStr = argsJson.startsWith("'") && argsJson.endsWith("'") ? argsJson.slice(1, -1) : argsJson
+
+    const parsed = JSON.parse(jsonStr)
     if (typeof parsed !== "object" || parsed === null) {
       throw new Error("Arguments must be a JSON object")
     }
@@ -85,7 +90,7 @@ function parseArguments(argsJson: string | Record<string, unknown> | undefined):
     throw new Error(
       `Invalid arguments JSON: ${errorMessage}\n\n` +
         `Expected a valid JSON object, e.g.: '{"key": "value"}'\n` +
-      `Received: ${argsJson}`
+        `Received: ${argsJson}`,
     )
   }
 }
@@ -95,10 +100,8 @@ export function applyGrepFilter(output: string, pattern: string | undefined): st
   try {
     const regex = new RegExp(pattern, "i")
     const lines = output.split("\n")
-    const filtered = lines.filter(line => regex.test(line))
+    const filtered = lines.filter((line) => regex.test(line))
-    return filtered.length > 0 
+    return filtered.length > 0 ? filtered.join("\n") : `[grep] No lines matched pattern: ${pattern}`
-      ? filtered.join("\n") 
-      : `[grep] No lines matched pattern: ${pattern}`
   } catch {
     return output
   }
@@ -114,8 +117,14 @@ export function createSkillMcpTool(options: SkillMcpToolOptions): ToolDefinition
       tool_name: tool.schema.string().optional().describe("MCP tool to call"),
       resource_name: tool.schema.string().optional().describe("MCP resource URI to read"),
       prompt_name: tool.schema.string().optional().describe("MCP prompt to get"),
-      arguments: tool.schema.string().optional().describe("JSON string of arguments"),
+      arguments: tool.schema
-      grep: tool.schema.string().optional().describe("Regex pattern to filter output lines (only matching lines returned)"),
+        .union([tool.schema.string(), tool.schema.record(tool.schema.string(), tool.schema.unknown())])
+        .optional()
+        .describe("JSON string or object of arguments"),
+      grep: tool.schema
+        .string()
+        .optional()
+        .describe("Regex pattern to filter output lines (only matching lines returned)"),
     },
     async execute(args: SkillMcpArgs) {
       const operation = validateOperationParams(args)
@@ -126,8 +135,9 @@ export function createSkillMcpTool(options: SkillMcpToolOptions): ToolDefinition
         throw new Error(
           `MCP server "${args.mcp_name}" not found.\n\n` +
             `Available MCP servers in loaded skills:\n` +
-          formatAvailableMcps(skills) + `\n\n` +
+            formatAvailableMcps(skills) +
-          `Hint: Load the skill first using the 'skill' tool, then call skill_mcp.`
+            `\n\n` +
+            `Hint: Load the skill first using the 'skill' tool, then call skill_mcp.`,
         )
       }
 

src/tools/skill/types.ts

@@ -18,7 +18,7 @@ export interface SkillInfo {
 }
 
 export interface SkillLoadOptions {
-  /** When true, only load from OpenCode paths (.opencode/skill/, ~/.config/opencode/skill/) */
+  /** When true, only load from OpenCode paths (.opencode/skills/, ~/.config/opencode/skills/) */
   opencodeOnly?: boolean
   /** Pre-merged skills to use instead of discovering */
   skills?: LoadedSkill[]