What problem does it solve? A single-pass plan review tends to rubber-stamp incomplete work: vague file references, missing acceptance gates, and unresolved design questions slip through. This Skill acts as an adversarial judge in the agent-quorum plan-refinement loop, independently deciding whether a plan is truly implementation-ready before a coding agent executes it. ## Core Features & Use Cases - Readiness verdicts: Returns a strict JSON verdict (ready true/false) conforming to readiness.schema.json, with no prose or markdown fences. - Invariant coverage accounting: Assesses every active invariant and retained occurrence exactly once with grounded evidence references, deriving coverage_complete and unresolved_occurrence_ids from exact accounting. - Scoped evaluation: Handles both intermediate revisions (with a single fixable revision_issue) and final canonical plans (independent verification where critique is advisory only). - Use Case: In a multi-agent plan → critique → update loop, run the judge after each revision to decide whether the exact plan revision has concrete file lists, per-phase acceptance gates, and no open blocker or major concerns before execution begins. ## Quick Start Evaluate this implementation plan for readiness and return the JSON verdict conforming to readiness.schema.json.
