What problem does it solve? Acceptance testing often stalls because failures go undiagnosed and unrouted, leaving teams unsure whether the spec, the code, or the test itself is wrong. This Skill runs the acceptance and live-integration rows from a test case catalog, records every result, and classifies each failure so the SDLC loop keeps moving. ## Core Features & Use Cases - Catalog-driven test implementation: Generates one acceptance test per catalog row, binding directly to .feature scenarios where a BDD runner exists so the specification is executed rather than transcribed. - Recorded execution: Runs the configured acceptance command and writes per-test outcomes, exit codes, durations, and failure output to an append-only run history under .agent/state/runs/. - Failure triage and routing: Classifies every failure as an implementation defect, test defect, spec gap, product gap, design gap, or flake, and routes it back to the responsible phase (code, tests, SRS, PRD, or techdoc). - Gate evaluation: Evaluates gate G5b, requiring recorded eval runs for AI features, and updates quality state with verdicts and counts. - Use Case: After unit tests pass for a feature, invoke /run-test FEAT-12 to execute its acceptance scenarios, get a triaged findings list, and know exactly which phase must fix each failure. ## Quick Start Run the acceptance test suite for feature FEAT-12 by invoking /run-test FEAT-12 and review the generated run report and triaged findings.
