| name | planner |
| description | Create step-by-step plans for solving problems, debugging issues, building features, or architecting complex solutions. Always includes pros/cons analysis and risk consideration. Use when the user asks to plan, create a roadmap, break down a task, design an approach, outline steps, or needs a structured path forward. |
Planner
Create clear, actionable step-by-step plans. Every plan considers trade-offs, risks, and alternatives so you make informed decisions before writing a single line of code.
How to Use
When the user asks to plan:
- Clarify the goal - Restate the objective in one sentence. Identify the plan type.
- Gather context - Examine relevant code, docs, configs, and constraints before planning.
- Identify approaches - For non-trivial problems, outline 2-3 candidate approaches before committing to one.
- Build the plan - Break the chosen approach into concrete, ordered steps.
- Analyze trade-offs - Add pros/cons for each significant decision point.
- Flag risks - Call out what could go wrong and how to mitigate it.
- Present the plan - Use the output format below.
Plan Types
Select the appropriate planning mode based on the user's goal:
Problem Solving
Break a well-defined problem into steps toward a solution.
- Focus on: Root cause, constraints, solution options, validation criteria
Debugging / Investigation
Systematically narrow down the source of a bug or unexpected behavior.
- Focus on: Hypotheses, evidence gathering, isolation steps, verification
Feature Building
Plan the implementation of a new feature within an existing codebase.
- Focus on: Requirements, affected components, data flow, backward compatibility, testing
Application / Architecture
Design a system or application from scratch or plan a major restructuring.
- Focus on: Component boundaries, technology choices, data model, integration points, scalability
Migration / Refactoring
Plan a safe transition from current state to desired state.
- Focus on: Incremental steps, rollback strategy, backward compatibility, zero-downtime
Planning Principles
- Start from what exists - Read the codebase before proposing changes. Reuse existing patterns.
- Smallest steps possible - Each step should be independently testable or verifiable.
- Dependencies first - Order steps so blockers are resolved early.
- Always have a rollback - For risky steps, define how to undo.
- Make trade-offs explicit - Never silently pick a side. Surface the decision.
- Scope ruthlessly - Separate "must have" from "nice to have". Flag scope creep.
Output Format
# Plan: [Goal in one sentence]
## Context
[Brief summary of current state, constraints, and relevant findings from codebase exploration]
## Approach
### Option A: [Name]
[One-paragraph description]
| Pros | Cons |
|------|------|
| ... | ... |
| ... | ... |
### Option B: [Name]
[One-paragraph description]
| Pros | Cons |
|------|------|
| ... | ... |
| ... | ... |
**Recommended**: Option [X] — [One-sentence justification]
## Steps
### Phase 1: [Phase Name]
- [ ] **Step 1**: [Action] — [Why this step matters]
- [ ] **Step 2**: [Action] — [Why this step matters]
- Verify: [How to confirm this step succeeded]
### Phase 2: [Phase Name]
- [ ] **Step 3**: [Action] — [Why this step matters]
- [ ] **Step 4**: [Action] — [Why this step matters]
- Verify: [How to confirm this step succeeded]
## Risks & Mitigations
| Risk | Likelihood | Impact | Mitigation |
|------|-----------|--------|------------|
| ... | Low/Med/High | Low/Med/High | ... |
| ... | Low/Med/High | Low/Med/High | ... |
## Out of Scope
- [Items explicitly excluded and why]
## Definition of Done
- [Criteria that confirm the plan is fully executed]
Debugging Plan Format
For debugging/investigation, use this specialized format instead:
# Investigation: [Symptom / Issue]
## Observed Behavior
[What is happening]
## Expected Behavior
[What should happen]
## Hypotheses
| # | Hypothesis | Likelihood | Evidence Needed |
|---|-----------|-----------|-----------------|
| 1 | ... | High/Med/Low | ... |
| 2 | ... | High/Med/Low | ... |
## Investigation Steps
- [ ] **Step 1**: [Check/test] — Tests hypothesis #[N]
- If confirmed: [Next action]
- If ruled out: Proceed to step [N]
- [ ] **Step 2**: [Check/test] — Tests hypothesis #[N]
- If confirmed: [Next action]
- If ruled out: Proceed to step [N]
## Resolution
[To be filled after investigation — root cause + fix]
Companion Skills
The planner works well in combination with other skills:
- Brainstormer (
/brainstormer) - Use before planning to generate diverse solution options when the path forward is unclear. Feed the top picks into the planner's "Approach" section.
- Challenger (
/challenger) - Use after planning to stress-test the plan. Feed the challenge report back to refine risks, mitigations, and step ordering.
Recommended workflow for complex problems:
/brainstormer — Generate options
/planner — Structure the best options into a concrete plan
/challenger — Poke holes in the plan
- Revise the plan based on challenge findings
Guidelines
- Keep steps atomic — one action per step, not "set up auth and add rate limiting".
- Include verification checkpoints after risky or complex steps.
- For plans longer than 10 steps, group into phases with clear milestones.
- When estimating effort, use T-shirt sizes (XS/S/M/L/XL) rather than hours.
- If the plan depends on unknowns, add a spike/discovery step early to resolve them.
- Prefer incremental delivery — the plan should produce working software at each phase boundary.
- Always list what is out of scope to prevent scope creep.