What problem does it solve? Vague or subjective success criteria cause QA loops that never close. This Skill converts any plan, spec, or task list into a structured YAML checklist where every requirement is either automatically testable (run a command, check output) or code-auditable (read specific files, verify logic), giving QA agents a precise target. ## Core Features & Use Cases - Requirement Classification: Splits plan requirements into automated checks (curl, build commands, test suites) and code-audit checks (UI logic, event handlers, business rules), maximizing automated coverage. - Structured YAML Output: Writes docs/qa-checklist.yaml with explicit comparison modes (exact, regex, numeric, exit_code), side-effect flags, stable category-based IDs, and status tracking for the QA checklist loop. - Incremental Updates: Auto-triggers on scope changes, bug reports, or feature feedback, updating only affected checks while preserving existing IDs and archiving removed ones. - Use Case: After writing a feature plan for a new checkout flow, run the Skill to produce a checklist covering every endpoint, UI state, and build gate, then hand it to the QA loop which iterates until all checks pass. ## Quick Start Ask the agent to generate success criteria from your plan file, for example: run /success-criteria docs/checkout-plan.md to produce docs/qa-checklist.yaml.
