7.0 KiB
description, argument-hint
| description | argument-hint |
|---|---|
| Restate requirements, assess risks, and create step-by-step implementation plan. WAIT for user CONFIRM before touching any code. | [feature description | path/to/*.prd.md] |
Plan Command
This command creates a comprehensive implementation plan before writing any code. It accepts either free-form requirements or a PRD markdown file.
Run inline by default. Do not call the Task tool or any subagent by default. This keeps /plan usable from plugin installs that ship commands without agent files.
What This Command Does
- Restate Requirements - Clarify what needs to be built
- Identify Risks - Surface potential issues and blockers
- Create Step Plan - Break down implementation into phases
- Wait for Confirmation - MUST receive user approval before proceeding
When to Use
Use /plan when:
- Starting a new feature
- Making significant architectural changes
- Working on complex refactoring
- Multiple files/components will be affected
- Requirements are unclear or ambiguous
How It Works
The assistant will:
- Analyze the request and restate requirements in clear terms
- Ground the plan in relevant codebase patterns when the repo is available
- Break down into phases with specific, actionable steps
- Identify dependencies between components
- Assess risks and potential blockers
- Estimate complexity (High/Medium/Low)
- Present the plan and WAIT for your explicit confirmation
Input Modes
| Input | Mode | Behavior |
|---|---|---|
path/to/name.prd.md |
PRD artifact mode | Read the PRD, pick the next pending delivery milestone or implementation phase, and write .claude/plans/{name}.plan.md |
| Any other markdown path | Reference mode | Read the file as context and produce an inline plan |
| Free-form text | Conversational mode | Produce an inline plan |
| Empty input | Clarification mode | Ask what should be planned |
In PRD artifact mode, create .claude/plans/ if needed. If the PRD contains a Delivery Milestones table, update only the selected row from pending to in-progress and set its Plan cell to the generated plan path. If the PRD uses the legacy .claude/PRPs/prds/ format with Implementation Phases, read it without migrating paths.
Pattern Grounding
Before writing the plan, search the codebase for conventions the implementation should mirror. Capture the top example for each relevant category with file references:
| Category | What to capture |
|---|---|
| Naming | File, function, type, command, or script naming in the affected area |
| Error handling | How failures are raised, returned, logged, or handled gracefully |
| Logging | Levels, format, and what gets logged |
| Data access | Repository, service, query, or filesystem patterns |
| Tests | Test file location, framework, fixtures, and assertion style |
If no similar code exists, state that explicitly. Do not invent a pattern.
PRD Artifact Output
When called with a .prd.md file, write the plan to .claude/plans/{kebab-case-name}.plan.md using this structure:
# Plan: {Feature Name}
**Source PRD**: {path}
**Selected Milestone**: {milestone or phase name}
**Complexity**: {Small | Medium | Large}
## Summary
{2-3 sentences}
## Patterns to Mirror
| Category | Source | Pattern |
|---|---|---|
| Naming | `path:line` | {short description} |
| Errors | `path:line` | {short description} |
| Tests | `path:line` | {short description} |
## Files to Change
| File | Action | Why |
|---|---|---|
| `path` | CREATE / UPDATE / DELETE | {reason} |
## Tasks
### Task 1: {name}
- **Action**: {what to do}
- **Mirror**: {pattern to follow}
- **Validate**: {command that proves correctness}
## Validation
```bash
{project-specific validation commands}
```
## Risks
| Risk | Likelihood | Mitigation |
|---|---|---|
## Acceptance
- [ ] All tasks complete
- [ ] Validation passes
- [ ] Patterns mirrored, not reinvented
After writing the artifact, report its path and WAIT for confirmation before writing code.
Example Usage
User: /plan I need to add real-time notifications when markets resolve
Assistant:
# Implementation Plan: Real-Time Market Resolution Notifications
## Requirements Restatement
- Send notifications to users when markets they're watching resolve
- Support multiple notification channels (in-app, email, webhook)
- Ensure notifications are delivered reliably
- Include market outcome and user's position result
## Implementation Phases
### Phase 1: Database Schema
- Add notifications table with columns: id, user_id, market_id, type, status, created_at
- Add user_notification_preferences table for channel preferences
- Create indexes on user_id and market_id for performance
### Phase 2: Notification Service
- Create notification service in lib/notifications.ts
- Implement notification queue using BullMQ/Redis
- Add retry logic for failed deliveries
- Create notification templates
### Phase 3: Integration Points
- Hook into market resolution logic (when status changes to "resolved")
- Query all users with positions in market
- Enqueue notifications for each user
### Phase 4: Frontend Components
- Create NotificationBell component in header
- Add NotificationList modal
- Implement real-time updates via Supabase subscriptions
- Add notification preferences page
## Dependencies
- Redis (for queue)
- Email service (SendGrid/Resend)
- Supabase real-time subscriptions
## Risks
- HIGH: Email deliverability (SPF/DKIM required)
- MEDIUM: Performance with 1000+ users per market
- MEDIUM: Notification spam if markets resolve frequently
- LOW: Real-time subscription overhead
## Estimated Complexity: MEDIUM
- Backend: 4-6 hours
- Frontend: 3-4 hours
- Testing: 2-3 hours
- Total: 9-13 hours
**WAITING FOR CONFIRMATION**: Proceed with this plan? (yes/no/modify)
Important Notes
CRITICAL: This command will NOT write any code until you explicitly confirm the plan with "yes" or "proceed" or similar affirmative response.
If you want changes, respond with:
- "modify: [your changes]"
- "different approach: [alternative]"
- "skip phase 2 and do phase 3 first"
Integration with Other Commands
After planning:
- Use the
tdd-workflowskill to implement with test-driven development - Use
/build-fixif build errors occur - Use
/code-reviewto review completed implementation - Use
/pror/prp-prto open a pull request
Need requirements first? Use
/plan-prdfor a lean PRD at.claude/prds/{name}.prd.md.Need the legacy PRP flow? Use
/prp-planfor deep PRP planning with.claude/PRPs/artifacts. Use/prp-implementto execute those plans with rigorous validation loops.
Optional Planner Agent
ECC also provides a planner agent for manual installs that include agent files. Use it only when the local runtime already exposes that subagent and the user explicitly asks you to delegate planning.
If the planner subagent is unavailable, continue planning inline instead of surfacing an "Agent type 'planner' not found" error.
For manual installs, the source file lives at:
agents/planner.md