What problem does it solve? Teams declare work "done" based on reviewer impression rather than agreed, checkable criteria, so review quality varies by who is looking and when. This Skill turns "done" into an objective, falsifiable, evidence-producing checklist designed from the project's actual surfaces, and separates what CI gates prove from what only human review can verify. ## Core Features & Use Cases - DoD design from scratch: Derives criteria from what the project actually delivers, tests each for objectivity, falsifiability, and evidence production, and names the four failure modes (vague criteria, unchecked lists, copied templates, over-heavy checklists). - Concrete nine-row DoD with a gate seam table: Applies this loop's own Definition of Done, marking which criteria CI proves (coverage, lint), which are partial (regressions, security posture), and which no gate can prove (observability, docs, consumer reach). - Use Case: A reviewer finishing a merge request walks the nine criteria, names the artifact a feature must reach outside the diff, and records which rows a green pipeline did not establish, instead of treating "CI is green" as "the DoD is met". ## Quick Start Ask the agent to evaluate whether the current slice satisfies the Definition of Done criteria and identify which rows no gate proves.
