refactor(athena): remove type assertions and improve agent factories

- Replace 'as AgentConfig' casts with proper typing in agent.ts and council-member-agent.ts
- Extract permission into typed variable following Sisyphus pattern
- Add GPT/non-GPT model branching to council-member-agent
- Use parseModelString for schema validation instead of inline logic
- Add strict() to council and athena config schemas
- Fix athena restriction list (remove redundant athena_council deny)
- Add orchestrator logging for council execution
- Update system prompt to notification-based workflow

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>

src/agents/athena/agent.ts

@@ -78,10 +78,13 @@ Step 3: Call athena_council ONCE PER MEMBER (member per tool call):
 - Launch all selected members first (one athena_council call per member) so they run in parallel.
 - Track every returned task_id and member mapping.
 
-Step 4: Collect all member outputs after launch:
-- For each tracked task_id, call background_output with block=true.
-- Gather each member's final output/status.
-- Do not proceed until every launched member has reached a terminal status (completed, error, cancelled, timeout).
+Step 4: Wait for all member outputs (DO NOT POLL):
+- After launching all members, STOP generating and end your turn. Do NOT call background_output yet.
+- The system will automatically notify you when tasks complete via system-reminder messages.
+- You will receive individual "[BACKGROUND TASK COMPLETED]" notifications as each member finishes.
+- When all members finish, you will receive an "[ALL BACKGROUND TASKS COMPLETE]" notification listing all task IDs.
+- ONLY after receiving the all-complete notification, call background_output(task_id="<id>") ONCE per member to retrieve their final output.
+- Do NOT loop or repeatedly call background_output. One call per task_id, only after the system notification.
 - Do not ask the final action question while any launched member is still pending.
 - Do not present interim synthesis from partial results. Wait for all members first.
 
@@ -115,7 +118,7 @@ The switch_agent tool switches the active agent. After you call it, end your res
 ## Constraints
 - Use the Question tool for member selection BEFORE calling athena_council (unless user pre-specified).
 - Use the Question tool for action selection AFTER synthesis (unless user already stated intent).
-- Use background_output (block=true) to collect member outputs after launch.
+- Do NOT poll background_output in a loop. Wait for system notifications, then collect once per member.
 - Do NOT call athena_council with multiple members in one call.
 - Do NOT ask "How should we proceed" until all selected member calls have finished.
 - Do NOT present or summarize partial council findings while any selected member is still running.
@@ -127,21 +130,26 @@ The switch_agent tool switches the active agent. After you call it, end your res
 export function createAthenaAgent(model: string): AgentConfig {
   const restrictions = createAgentToolRestrictions(["write", "edit"])
 
+  const permission = {
+    ...restrictions.permission,
+    question: "allow",
+  } as AgentConfig["permission"]
+
   const base = {
     description:
-      "Primary synthesis strategist for multi-model council outputs. Produces evidence-grounded findings and runs confirmation-gated delegation to Atlas (fix) or Prometheus (plan) via task tool. (Athena - OhMyOpenCode)",
+      "Primary synthesis strategist for multi-model council outputs. Produces evidence-grounded findings and runs confirmation-gated delegation to Atlas (fix) or Prometheus (plan) via switch_agent. (Athena - OhMyOpenCode)",
     mode: MODE,
     model,
     temperature: 0.1,
-    permission: { ...restrictions.permission, question: "allow" as const },
+    permission,
     prompt: ATHENA_SYSTEM_PROMPT,
     color: "#1F8EFA",
-  } as AgentConfig
-
-  if (isGptModel(model)) {
-    return { ...base, reasoningEffort: "medium" } as AgentConfig
   }
 
-  return { ...base, thinking: { type: "enabled", budgetTokens: 32000 } } as AgentConfig
+  if (isGptModel(model)) {
+    return { ...base, reasoningEffort: "medium" }
+  }
+
+  return { ...base, thinking: { type: "enabled", budgetTokens: 32000 } }
 }
 createAthenaAgent.mode = MODE

src/agents/athena/council-member-agent.ts

@@ -1,5 +1,6 @@
 import type { AgentConfig } from "@opencode-ai/sdk"
 import type { AgentMode } from "../types"
+import { isGptModel } from "../types"
 import { createAgentToolRestrictions } from "../../shared/permission-compat"
 
 const MODE: AgentMode = "subagent"
@@ -16,13 +17,20 @@ export function createCouncilMemberAgent(model: string): AgentConfig {
     "athena_council",
   ])
 
-  return {
-    description: "Independent code analyst for Athena multi-model council. Read-only, evidence-based analysis. (Council Member - OhMyOpenCode)",
+  const base = {
+    description:
+      "Independent code analyst for Athena multi-model council. Read-only, evidence-based analysis. (Council Member - OhMyOpenCode)",
     mode: MODE,
     model,
     temperature: 0.1,
     prompt: COUNCIL_MEMBER_PROMPT,
     ...restrictions,
-  } as AgentConfig
+  }
+
+  if (isGptModel(model)) {
+    return { ...base, reasoningEffort: "medium" }
+  }
+
+  return { ...base, thinking: { type: "enabled", budgetTokens: 32000 } }
 }
 createCouncilMemberAgent.mode = MODE

src/agents/athena/council-orchestrator.ts

@@ -1,4 +1,5 @@
 import type { LaunchInput, BackgroundTask } from "../../features/background-agent/types"
+import { log } from "../../shared/logger"
 import { buildCouncilPrompt } from "./council-prompt"
 import { parseModelString } from "./model-parser"
 import type { CouncilConfig, CouncilLaunchFailure, CouncilLaunchedMember, CouncilLaunchResult, CouncilMemberConfig } from "./types"
@@ -25,6 +26,9 @@ export interface CouncilExecutionInput {
  */
 export async function executeCouncil(input: CouncilExecutionInput): Promise<CouncilLaunchResult> {
   const { question, council, launcher, parentSessionID, parentMessageID, parentAgent } = input
+
+  log("[athena-council] Executing council", { memberCount: council.members.length })
+
   const prompt = buildCouncilPrompt(question)
 
   const launchResults = await Promise.allSettled(
@@ -50,6 +54,8 @@ export async function executeCouncil(input: CouncilExecutionInput): Promise<Coun
     })
   })
 
+  log("[athena-council] Council execution complete", { launched: launched.length, failures: failures.length })
+
   return {
     question,
     launched,

src/agents/athena/index.ts

@@ -4,4 +4,4 @@ export * from "./council-member-agent"
 export * from "./model-parser"
 export * from "./council-prompt"
 export * from "./council-orchestrator"
-export * from "../../config/schema/athena"
+

src/config/schema/athena.ts

@@ -1,14 +1,12 @@
 import { z } from "zod"
+import { parseModelString } from "../../agents/athena/model-parser"
 
 /** Validates model string format: "provider/model-id" (e.g., "openai/gpt-5.3-codex"). */
 const ModelStringSchema = z
   .string()
   .min(1)
   .refine(
-    (model) => {
-      const slashIndex = model.indexOf("/")
-      return slashIndex > 0 && slashIndex < model.length - 1
-    },
+    (model) => parseModelString(model) !== null,
     { message: 'Model must be in "provider/model-id" format (e.g., "openai/gpt-5.3-codex")' }
   )
 
@@ -20,9 +18,9 @@ export const CouncilMemberSchema = z.object({
 
 export const CouncilConfigSchema = z.object({
   members: z.array(CouncilMemberSchema).min(2),
-})
+}).strict()
 
 export const AthenaConfigSchema = z.object({
   model: z.string().optional(),
   council: CouncilConfigSchema,
-})
+}).strict()

src/shared/agent-tool-restrictions.ts

@@ -14,7 +14,7 @@ const EXPLORATION_AGENT_DENYLIST: Record<string, boolean> = {
 }
 
 const ATHENA_RESTRICTIONS = permissionToToolBooleans(
-  createAgentToolRestrictions(["write", "edit", "athena_council"]).permission
+  createAgentToolRestrictions(["write", "edit"]).permission
 )
 
 const AGENT_RESTRICTIONS: Record<string, Record<string, boolean>> = {