---

name: testing-qa description: "Select LangBot verification layers, run focused pytest, pnpm, Box, startup, and lbs QA checks, interpret fixtures and manual_check readiness, and collect evidence without over-running slow gates." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Testing and QA

Use this sub-skill when the task asks which LangBot tests to run, how to verify a change, how to use the in-repo skills/ QA assets, or how to interpret fixture/manual readiness and evidence requirements.

Read First

• references/test-selection.md for layered gates and change-surface-to-command mapping.
• references/lbs-qa-assets.md for skills/bin/lbs, cases, suites, fixtures, troubleshooting, and UI evidence rules.
• references/troubleshooting.md for flaky, skipped, service-backed, frontend, and browser/fixture failures.

Golden Rule

Run the narrowest meaningful check first. Expand only when the change crosses subsystems or the focused test is too synthetic for the risk.

Command Selector

python scripts/select_langbot_checks.py quick
python scripts/select_langbot_checks.py api-mcp
python scripts/select_langbot_checks.py pipeline
python scripts/select_langbot_checks.py plugin-box
python scripts/select_langbot_checks.py persistence-rag
python scripts/select_langbot_checks.py frontend
python scripts/select_langbot_checks.py skills-lbs

Add --run --repo-root /path/to/LangBot only after prerequisites are confirmed.

Evidence Discipline

• Do not turn optional backend skips into passes.
• UI/browser test cases require UI evidence; API/curl checks are diagnostic for UI paths, not replacements.
• Preserve command, scope, result, skipped prerequisites, and artifacts.
• For live provider/platform/browser tests, state credentials/preconditions and avoid printing secrets.