smithery/metabench

targeted-testing

Pick and run the smallest correct validation step (checks → focused Jest by path → broader suites). Use whenever you modify code and need confidence quickly.

Installation

$ npx skills add smithery/metabench --skill targeted-testing

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from smithery/metabench.

npx skills add smithery/metabench

Browse all from smithery/metabench

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 1,924 B
  • docs SUMMARY.md 185 B

History

  1. First recorded snapshot · 0 installs

SKILL.md

Targeted Testing

Scope

  • Select the smallest test/check that proves the change
  • Prevent slow, flaky, or irrelevant test runs
  • Capture evidence in session notes

Inputs

  • Files changed
  • Whether the change is UI server / CLI / DB adapter / crawler
  • Existing check scripts near the code

Procedure

  1. Require smoke test — the cheapest possible validation. Before running any test suite, verify the module loads without errors:

``bash node -e "require('./path/to/module'); console.log('OK')" `` This catches missing imports, syntax errors, and circular dependencies in <1 second. Especially valuable after decomposing a monolith into smaller modules — smoke-test each extracted module individually.

  1. Prefer a local check script first (fast, deterministic, exits cleanly).
  2. Then run the smallest Jest suite by path.
  3. Only then widen to broader suites if needed.
  4. Record the exact commands and outcomes in WORKING_NOTES.md.

If you need browser semantics but don’t want full Jest E2E yet, prefer the single-browser scenario runner:

  • node tools/dev/ui-scenario-suite.js --suite=scripts/ui/scenarios/<suite>.suite.js --scenario=001 --print-logs-on-failure

Validation

  • Official Jest invocation rule (repo policy): use npm run test:by-path <file>.

Escalation / Research request

Ask for dedicated research if:

  • there is no obvious “smallest check” and you need a new check harness
  • the feature spans multiple systems and you need a validation ladder proposal

References

  • Testing quick reference: docs/TESTINGQUICKREFERENCE.md
  • Validation ladder guidance: AGENTS.md
  • WebSocket upgrade skill: docs/agi/skills/websocket-upgrade/SKILL.md