What problem does it solve? Running a brainstorm and then a separate dual-reviewer review of a change idea normally requires two manual command invocations with a stop in between. This skill removes that seam by running the brainstorm phase and the main-agent plus reviewer-agent review loop in a single orchestrated flow, passing the change id explicitly between phases. ## Core Features & Use Cases - Two-phase orchestration: Phase A produces a decision capsule at openspec/changes/<change-id>/brainstorm.md; Phase B runs the dual-reviewer (default Claude + Codex) inline-fix review loop without re-confirming scope. - Brainstorm-gate safety: If Phase A fails to write brainstorm.md, the skill stops before Phase B and reports how to recover, instead of reviewing a missing artifact. - Pre-resolved inputs: Consumes an already-allocated change id, a pre-resolved codex.mode decision, and a resolved model target, so it never re-runs branch guards, mode resolution, or id allocation. - Use Case: You describe a feature in one line via /ptp:brainstorm-full; the skill brainstorms alternatives, records the decision capsule, then runs Claude and Codex review loops and reports a terminal state such as BOTH PHASES DONE with a review tally and the next-step recommendation /ptp:plan <change-id>. ## Quick Start Run the /ptp:brainstorm-full command with a one-line description of your change to brainstorm it and immediately dual-review the resulting brainstorm document.
