What problem does it solve? A thin or hand-wavy brainstorm silently produces thin OpenSpec proposal and spec artifacts downstream. This Skill catches and fixes brainstorm weaknesses before any artifacts exist by reviewing the change's brainstorm.md with two independent reviewers and editing it inline until both loops converge. ## Core Features & Use Cases - Dual-reviewer convergence loop: Runs the main agent's review loop first, then a gated reviewer-agent loop (Codex by default), each iterating review, confirm, and inline fix until zero confirmed findings or the configured iteration cap. - Inline targeted fixes: Resolves confirmed findings with minimal edits to brainstorm.md, never regenerating the brainstorm and never running openspec validate. - Combined terminal state and marker: Folds both phases into one terminal state (e.g. BOTH PHASES DONE) and performs a single atomic write of the stages/brainstorm.json convergence marker. - Use Case: After drafting a brainstorm for a new feature change, run the full brainstorm review so Claude and Codex independently audit it, fix gaps like missing options or undocumented assumptions, and confirm it is sound before authoring OpenSpec artifacts. ## Quick Start Ask the agent to run /ptp:review-brainstorm-full on your change id to dual-review and inline-fix its brainstorm.md before planning.
