testing-qa

Router skill. Resolve topic, then Read EXACTLY ONE playbook from routes:. [DOC]

When to use

• A request names a test type (unit, E2E/BDD, cross-browser, performance) or asks for a test plan/coverage gate. Single dimension → that topic; "what should we test / where to invest" → test-strategy. [INFERENCE]
• Skip if the ask is non-testing (build, deploy, lint) — route elsewhere. [INFERENCE]

Routing

1. Infer topic from the request; ask only on genuine ambiguity between two enums. [DOC]
2. Map topic → routes: path, Read that ONE file. Never load the cluster. [DOC]
3. depth=deep → apply the playbook exhaustively, verifying at each step; quick → essentials only (the path most requests need). [DOC]

Disambiguation (overlapping asks)

• BDD scenarios driving a browser → bdd-full-spectrum (spec-first), not e2e-testing. [INFERENCE]
• Load/latency/throughput → performance-testing, even if framed as "E2E under load". [INFERENCE]
• "Increase coverage" with no type named → unit-testing (coverage gate lives there). [INFERENCE]

Spine & gates

• Spine: Discover → Analyze → Execute → Validate. [DOC]
• Gates: constitution v6.0.0 (enforcement), evidence tags (Alfa core set, EN spelling per references/verification-tags.md), script-first rule. [DOC]

Acceptance criteria

• Exactly one playbook Read; topic set; depth respected. [DOC]
• Validate step run before "done"; no green-as-success without evidence. [DOC]
• Gate the deliverable with assets/ (routing checklist + quality rubric). [DOC]

Anti-patterns

• Reading 2+ playbooks "to compare", or summarizing a topic from memory instead of its playbook. [INFERENCE]
• Guessing topic silently when two enums fit equally — ask. [INFERENCE]
• Marking complete on an unrun Validate step. [DOC]