discover-testing git:20261012.f546f4cA

Immutable. This exact content never changes and is served at /api/v1/blob/f97e7d5bdb40d207.

---
name: discover-testing
description: Use this skill when you need to route a request to the right testing skill before execution; triggers include discover testing, testing skill router, and which testing skill.
---

# Testing Skill Discovery (English)

**Chinese version:** See the corresponding Chinese skill.

## When to Use

- Need to decide which testing skill should be used before execution.
- The request mixes multiple testing directions or phases.

## Full inventory and availability

Read `references/skill-capability-inventory.json` to check exact canonical IDs. This 2026-10-12 snapshot catalogs 165 IDs; catalog existence does not prove installation or callability on the current Host. Missing installation evidence means availability=`UNKNOWN`; explicitly absent means `UNAVAILABLE`. Match explicit specialties (authentication vs authorization, retrieval vs answer quality, impact scope vs regression selection) before broad Composition routes. Keep one primary and at most one helper. For an unavailable target, state the gap and minimal installation/handoff action without claiming invocation.

## Workflow

1. Read the same-directory `reference.md` first and match one structured route when applicable; it is the only Composition reference required for independent installation.
2. Read the user request and first identify the capability stage (Core QA / Engineering QA / Production Quality / AI Native QA) and primary testing goal.
3. Follow the routing prompt under `prompts/`: pick 1 primary skill; add at most 1 helper only when needed.
4. Hand the request to the target skill; do not execute the full testing work inside this router skill.

## Core Constraints

- Recommend few skills — avoid menu dumping.
- If the target skill is already obvious, say so directly.
- Make the route actionable: name the skill and the reason.
- Prefer the Composition route's one-primary/one-optional constraint; relations are navigation metadata, not installation dependencies or a mandatory execution chain.
- Use `ai-assisted-testing` for AI for QA; Testing for AI belongs to AI Native QA. Until roadmap Skills are installed, never recommend them as callable primary Skills.

## Progressive Disclosure

- Before producing output, read and follow `prompts/discover-testing.md` (minimum coverage, output structure, quality bar).
- When Excel/CSV/JSON/Word is requested: read `output-formats.md` and honor the format.
- When a ready-made template fits: use matching files under `output-templates/`.
- For format conversion or helper checks: prefer existing `scripts/` over reinventing.
- For evaluating/regressing this skill: use `evals/` with skill-up.
- For step ↔ prompt mapping: keep using this package's `reference.md`; do not follow relative links into another Skill's internal files.

## Pre-delivery Checklist

- [ ] Followed the main prompt's output structure
- [ ] Minimum coverage focus: main goal, best-fit primary skill, optional supporting skill, why this choice fits, next step to continue work (details in main prompt)
- [ ] Covered the minimum checklist, or explained omissions
- [ ] High-risk items have explicit priority
- [ ] Did not invent details the user did not provide
- [ ] Assumptions and gaps are marked

## Common Pitfalls

- Do not recommend many skills at once.
- Do not turn skill selection into full test execution.
- Do not pretend a route is complete when information is insufficient.