What problem does it solve? Turning a feature request or bug fix into reviewed, mergeable code often skips planning discipline, producing unreviewed diffs and scope creep. This Skill enforces a structured workflow that converts user intent into a token-bounded specification, implements it, and runs multi-agent adversarial review before presenting results. ## Core Features & Use Cases - Intent Clarification & Routing: Clarifies the request, checks git state, detects multi-goal scope, and routes work to either a one-shot path or a full plan-code-review pipeline. - Spec-Driven Planning: Generates a frozen-intent specification with boundaries, I/O edge-case matrices, code maps, and Given/When/Then acceptance criteria, kept within a 900-1600 token budget. - Adversarial Review Loop: Launches three independent reviewers (blind hunter, edge-case hunter, acceptance auditor), classifies findings as intent_gap, bad_spec, patch, defer, or reject, and loops back when needed. - Use Case: A developer asks to add pagination to an API endpoint. The Skill drafts a spec, gets human approval, implements the change, runs three review passes, fixes patches, and presents a clickable Suggested Review Order for final human review. ## Quick Start Ask the agent to implement a specific feature or bug fix, for example: implement JWT refresh token support in the auth module following the project conventions.
