---
status: Planning
appetite: [Small: 1-2 days | Medium: 3-5 days | Large: 1-2 weeks]
owner: [Name]
created: [YYYY-MM-DD]
tracking: [GitHub Issue URL or Notion page URL - added automatically]
---

# [Feature Name]

## Problem

[Real scenario showing the pain. User perspective. Specific, not vague.]

**Current behavior:**
[What happens now that's broken/painful]

**Desired outcome:**
[What success looks like]

## Appetite

**Time budget:** [Small: 1-2 days | Medium: 3-5 days | Large: 1-2 weeks]

**Team size:** [Solo | Pair | Small team]

## Solution

### Key Elements

- **[Component 1]**: [What it does, not how]
- **[Component 2]**: [What it does, not how]
- **[Component 3]**: [What it does, not how]

### Flow

[Breadboard-style flow showing user journey]

**Starting point** → [Action/affordance] → **Next place** → [Action/affordance] → **End state**

Example:
Settings page → Click "Enable 2FA" → Setup screen → Enter code → Confirmation → Back to settings (with 2FA enabled)

### Technical Approach

[High-level technical direction - stay abstract enough for implementation flexibility]

- [Key decision 1]
- [Key decision 2]
- [Integration points]

## Rabbit Holes & Risks

### Risk 1: [Description]
**Impact:** [What breaks if this goes wrong]
**Mitigation:** [How we'll handle it]

### Risk 2: [Description]
**Impact:** [What breaks if this goes wrong]
**Mitigation:** [How we'll handle it]

[Add more risks as needed]

## No-Gos (Out of Scope)

[Explicitly state what we're NOT doing. This is critical for scope control.]

- [Feature deferred to later]
- [Edge case we'll handle in v2]
- [Related but separate concern]

## Success Criteria

[Measurable outcomes tied to the appetite. What does "done" look like?]

- [ ] [Criterion 1]
- [ ] [Criterion 2]
- [ ] [Criterion 3]

## Team Orchestration

When this plan is executed, the lead agent orchestrates work using Task tools. The lead NEVER builds directly - they deploy team members and coordinate.

### Team Members

[List each team member needed. Name them uniquely so they can be referenced in tasks.]

- **Builder ([component-name])**
  - Name: [unique-name, e.g., "api-builder"]
  - Role: [Single focused responsibility]
  - Agent Type: [builder | designer | tool-developer | database-architect | etc.]
  - Resume: true

- **Validator ([component-name])**
  - Name: [unique-name, e.g., "api-validator"]
  - Role: [What they verify]
  - Agent Type: validator
  - Resume: true

[Add more team members as needed. Pattern: builder + validator pairs for each major component.]

### Available Agent Types

**Builders:**
- `builder` - General implementation (default for most work)
- `designer` - UI/UX following design systems
- `tool-developer` - High-quality tool creation
- `database-architect` - Schema design, migrations
- `agent-architect` - Agent systems, context management
- `test-engineer` - Test implementation
- `documentarian` - Documentation updates
- `integration-specialist` - External service integration

**Validators:**
- `validator` - Read-only verification (no Write/Edit tools)
- `code-reviewer` - Code review, security checks
- `quality-auditor` - Standards compliance

**Service Agents:**
- `github`, `notion`, `linear`, `stripe`, `sentry`, `render`

## Step by Step Tasks

[Each task maps to a `TaskCreate` call. Execute top to bottom. Build tasks can run in parallel; validators wait for their builder.]

### 1. [First Build Task]
- **Task ID**: build-[component]
- **Depends On**: none
- **Assigned To**: [builder name from Team Members]
- **Agent Type**: [agent type]
- **Parallel**: true
- [Specific action to complete]
- [Specific action to complete]

### 2. [Validation Task]
- **Task ID**: validate-[component]
- **Depends On**: build-[component]
- **Assigned To**: [validator name from Team Members]
- **Agent Type**: validator
- **Parallel**: false
- Verify implementation meets criteria
- Run validation commands
- Report pass/fail status

[Continue pattern for each component...]

### N. Final Validation
- **Task ID**: validate-all
- **Depends On**: [all previous task IDs, comma-separated]
- **Assigned To**: [lead validator]
- **Agent Type**: validator
- **Parallel**: false
- Run all validation commands
- Verify all success criteria met
- Generate final report

## Validation Commands

[Commands to verify the work is complete - used by validators]

- `[command 1]` - [what it validates]
- `[command 2]` - [what it validates]

---

## Open Questions

[Critical unknowns that need supervisor input before finalizing]

1. [Question about scope/approach]
2. [Question about priority/tradeoff]
3. [Question about technical constraint]

gh issue create \
  --repo {org}/{repo} \
  --title "[Plan] {Feature Name}" \
  --label "plan" \
  --body "$(cat <<'EOF'
## Plan Document

See: docs/plans/{slug}.md (branch: plan/{slug})

**Appetite:** {appetite}
**Status:** Planning

---
This issue tracks the plan at `docs/plans/{slug}.md`. Update the plan document for details; this issue is for tracking and discussion.
EOF
)"

Plan draft created: docs/plans/{slug}.md

GitHub URL: https://github.com/{org}/{repo}/blob/{branch}/docs/plans/{slug}.md
Tracking: {GitHub issue URL or Notion page URL}

I've made the following key assumptions:
- [Assumption 1]
- [Assumption 2]
- [Assumption 3]

Please review the Open Questions section at the end of the plan and provide answers so I can finalize it.

Plan finalized: docs/plans/{slug}.md

GitHub URL: https://github.com/tomcounsell/ai/blob/{branch}/docs/plans/{slug}.md

When you're ready to implement, use this prompt:

---
Implement the plan in docs/plans/{slug}.md

Follow the solution approach, stay within the appetite, and avoid the identified rabbit holes. Check off success criteria as you complete them.
---

I need to narrow this down. What specific problem are you trying to solve?

- Is login failing for certain users?
- Is the flow too complicated?
- Is there a security concern?
- Is it too slow?

Or is there a specific story: "Users with [X situation] can't [Y action] because [Z breakdown]"?

This feels like multiple separate features bundled together. Each of these could be its own plan:

1. Dark mode toggle
2. Notification preferences
3. Profile editing
4. Data export

Which one is most urgent? Or is there a single pain point driving all of these?

Great - this is specific. I'll create a plan for async meeting rescheduling.

Starting with Medium appetite (3-5 days) since it involves workflow changes and possibly notifications.