What problem does it solve?
Reviewing AI-agent-implemented work packages requires comparing code against specs, plans, and acceptance criteria, then recording a verdict in a traceable way. This Skill standardizes that review loop so approvals and rejections are captured through the deterministic event log rather than ad-hoc edits.
Core Features & Use Cases
- Structured WP Review: Compare implementation against WP scope, spec, plan, and acceptance checks before deciding.
- Approve/Reject Verdicts: Approve with
spec-kitty agent tasks move-task <WP> --to approved or reject with --to planned --review-feedback-file <f>.
- Authoritative Event Log: Records the
review_result event in status.events.jsonl as the sole source of truth, keeping the review-cycle-N.md render non-authoritative.
- Use Case: After an agent finishes a work package in an isolated git worktree, use this Skill to verify the implementation against acceptance criteria and either approve it for merge or reject it with concrete, actionable feedback.
Quick Start
Review the current work package against its spec and acceptance criteria, then approve it or reject it with feedback using the spec-kitty move-task command.