What problem does it solve? Jumping straight into coding a multi-file feature without a plan leads to missed edge cases, unclear scope, and rework. This Skill designs the feature before any code is written by having two subagents argue over the approach and producing a reviewed plan. ## Core Features & Use Cases - Dual-subagent design: Spawns a planner subagent to draft the plan and a skeptic subagent to attack it, then merges their outputs. - Memory Bank awareness: Reads .memory-bank/index.md for project context so plans reflect existing architecture. - Structured plan output: Writes swarm-report/<slug>-plan.md with acceptance criteria, affected files, tests, blockers, and assumptions. - Use Case: Before implementing "add OAuth login to the API", run the Skill to get a plan with acceptance criteria that /code-review can later verify, plus a list of unresolved blockers for human decision. ## Quick Start Ask the AI to run /code-plan with a one-sentence description of the feature you want to build, such as "/code-plan add rate limiting to the public API endpoints".
