What problem does it solve? Per-slice tests prove each slice works in isolation, but nothing verifies that the assembled feature satisfies the PRD's epic-level Definition of Done. This Skill closes that gap with an adversarial integration and edge-case pass before an epic is allowed to ship. ## Core Features & Use Cases - Epic-level verification: Builds a rubric from the PRD (Definition of Done, goals, user stories, P0 requirements, NFRs, edge cases) and tests the assembled feature against it. - Two-scope adversarial pass: Hunts for failures both inward (integration seams between slices) and outward (blast-radius regressions in adjacent pre-existing code, mapped via /zoom-out). - Earned close-out: On pass, flips the PRD to Shipped, posts a closing comment, and closes the epic; on failure, posts an expected-vs-actual report and files full vertical-slice gap issues under the same PRD-ID. - Use Case: After grill -> to-prd -> to-issues -> /tdd completes all slices of PRD-042, run this Skill to confirm the whole feature works end-to-end and decide whether the epic can ship. ## Quick Start Run the close-epic skill on PRD-042 to verify the assembled epic against its Definition of Done and close it if it passes.
