mega-plan

# Mega-Plan Skill

Transform conversations into rock-solid implementation plans. The conversation IS the planning process.

<core_philosophy>
- **One output**: PLAN.md - no JSON, no state files, no orchestration overhead
- **Self-contained**: Implementer needs nothing beyond the plan
- **Research visible**: User sees findings, guides direction
- **Iterative**: User reviews drafts, suggests changes before finalizing
</core_philosophy>

<design_decisions>
## Questions That Force Decisions

Before exploring solutions, force these choices:

**Scope decisions**
- "What's the simplest version that still delivers value?" → Forces MVP boundary
- "If you could only ship one capability, which?" → Forces prioritization

**Quality decisions**
- "What would make this a failure even if it 'works'?" → Forces success criteria beyond "it runs"
- "Who's the handoff audience: you tomorrow, a teammate, or a future AI agent?" → Forces documentation depth

**Build vs integrate decisions**
- "Is this core to your product or commodity infrastructure?" → Forces build-vs-buy stance
- "What existing pattern in the codebase is this most similar to?" → Forces reuse consideration

Proceed to EXPLORE only after the user commits to specific answers. If responses remain vague ("it depends"), rephrase with concrete options until you receive a clear commitment.
</design_decisions>

<planning_anti_patterns>
## Patterns to Avoid

Plans fail when implementers are left guessing. Avoid these patterns:

**Vague action items**
- "Implement the feature" → Instead: "Create `src/services/feature.ts` with `handleX()` function following pattern in `src/services/auth.ts:45-60`"
- "Add proper error handling" → Instead: List the 3-5 specific errors, where caught, what happens for each

**Assumed context**
- "Use the standard approach" → Instead: Name the file containing the pattern and line numbers
- "Follow existing conventions" → Instead: Specify which file demonstrates the convention

**Pseudocode in research no