What problem does it solve? Verifying whether a ticket's technical acceptance criteria are actually implemented is error-prone when done by a single model pass or by trusting commit messages and ticket IDs. This Skill freezes the ticket text and repository state at fixed commit SHAs, then checks each criterion against reproducible ground truth so that unmet, contradicted, or unverifiable requirements are surfaced explicitly instead of being smoothed over. ## Core Features & Use Cases - Criterion-by-criterion review: Decomposes bundled acceptance criteria into atomic subclaims, classifies them (factual, judgment, systemic, negative, runtime), and routes each to isolated primary reviewer processes. - Evidence-over-majority adjudication: Mechanically re-verifies every cited file, line range, and test execution at the frozen commit; negative claims require multiple probes or remain UNKNOWN, and no majority vote replaces evidence. - Tracker integration with opt-in publishing: Reads tickets from Jira (acli/REST), YouTrack (MCP/REST), or GitLab (glab), renders a structured review with per-criterion dispositions (KEEP, KEEP_UNMET, REFINE, SPLIT, and more), and publishes comments only after explicit approval and readback. - Use Case: Before a QA handoff, run the review against a Jira ticket and two pinned repository commits to learn which acceptance criteria are demonstrably implemented, which are only partially evidenced, and which need refinement or a product decision. ## Quick Start Review the technical acceptance criteria of Jira ticket PROJ-123 against the current main branch commits of this repository and produce a consolidated review without publishing anything.
