What problem does it solve? Plans often contain unverified assumptions, overlap with existing code, and unscoped scope creep. This Skill stress-tests an existing plan file against the actual codebase, editing it in place and recording every finding so the plan converges before implementation begins. ## Core Features & Use Cases - Codebase-grounded review rounds: Challenges plan assumptions with concrete file:line evidence, checks overlap against existing utilities, and applies YAGNI and evidence-quality gates to every plan item. - Two review modes: Lightweight self-review for small plans, or team mode that dispatches specialist agents (security, concurrency, UX, data, and more) with domain-scoped briefs for medium and large plans. - Cross-referenced audit trail: Records every finding in review-findings.md and every round in review-iteration-history.md under the plan's artifacts/ folder, with a validation script that verifies cross-reference integrity. - Spec-aware mode: Detects feature-specification.md files and enforces behavioral-level language, routing implementation mechanics into a technical-notes companion file. - Use Case: After drafting a feature plan, run this Skill to have adversarial and domain specialists refute weak assumptions, flag missing failure modes, and tighten the plan across up to three rounds before writing any code. ## Quick Start Ask the AI to iteratively review and tighten the plan at ~/.claude/plans/my-feature.md using a medium-sized specialist team.
