What problem does it solve? AI coding agents frequently mark work as done without real evidence, leaving broken features and fabricated test results. This Skill enforces a two-tier validation checklist so a scope can only transition to Done when every Definition of Done item carries captured proof. ## Core Features & Use Cases - Tier 1 Universal Checks: Verifies all DoD checkboxes are checked, each with inline evidence of at least 10 lines of raw output, plus test plan parity, test file existence, zero warnings, and zero TODO/FIXME markers. - Tier 2 Role-Specific Profiles: Loads role-matched validation rules for implement, test, validate, audit, docs, and bug agents, such as adversarial regression tests for bug fixes. - Mode Ceiling Pre-Flight: Checks the workflow mode status ceiling before any non-spec file edit and returns a route_required result when the ceiling forbids the change. - Use Case: Before flipping a scope status to Done or emitting a completed_owned result envelope, run this validation to confirm every DoD item has genuine captured evidence rather than narrative claims. ## Quick Start Validate whether the current scope's Definition of Done items all pass the Tier 1 and Tier 2 checks before marking it Done.
