What problem does it solve? In multi-session, multi-agent software engineering, implementations can drift from their approved design and plan, and self-reported success claims are unreliable. This Skill acts as an independent acceptance reviewer that verifies a plan's implementation against the governing architecture, immutable plan, real runtime behavior, and downstream contracts before it is accepted or rejected. ## Core Features & Use Cases - Adversarial acceptance review: Builds a traceability matrix mapping every plan goal and acceptance checkbox to architecture sections, commits, and independent test evidence, classifying each as PASS, FAIL, UNVERIFIED, or NOT_APPLICABLE. - Evidence-based lifecycle transitions: Accepts or rejects plans through the MINE MCP tools or JSON CLI (mine plan accept / mine plan reject), never editing the execution graph files directly, and records review reports under docs/plan/reports/. - Scoped reviewer fixes vs. compensating plans: Classifies findings by impact, fixing narrow defects directly with separate reviewer commits, or rejecting and registering a compensating plan for substantial design-level issues. - Release closure mode: Verifies the final sync, runs the stale plan-reference scan (mine scan plan-refs --check), constructs the stable candidate tree, and completes the release closure gates. - Use Case: After an agent finishes implementing plan03, invoke the review to independently re-run its acceptance commands, inspect edge cases, merge the branch into dev with --no-ff, and transition the plan to ACCEPTED with a committed review report. ## Quick Start Ask the agent to review the implemented plan by invoking mine-plan-review with the plan ID, for example: review plan03 and tell me whether it can be accepted.
