Qa Skills

Key features

• "Help me test this requirement"QA orchestrates 9-stage pipeline, checkpoints are waiting for your judgment: full set of test assets + test report
• Use cases written by AI often look professional, but are actually unexecutable - with vague decisions, placeholders, no decision time limit, and fictitious entrances.
• By cramming the methodology, templates, and rules into one SKILL.md, the Agent can effectively follow fewer rules (Red Hat ACE practice summary: performance degrades after instructions exceed 500 lines).
• When formulating the testing strategy for the model without skill installed, in the 30 evaluation samples across the two model levels, no explicit type decision was given - performance and safety were "mentioned" in the output, but it was never decided which ones should be included, how deep they should be tested, and which ones should be explicitly not tested.
• "Write use cases based on this PRD": code first: request the warehouse, read the implementation, review potential bugs and then write: dual-track use cases: markmap (human) + schema.yaml (machine)