SituationalMITOriginal

Skill Tester

Dry-runs a skill against a test input to verify trigger detection, output shape and side-effect safety before you ship it.

Entry verified April 21, 2026

The short answer

The QA step for skill authors. Simulates invocation, checks that the skill's trigger matches the test input, and verifies the skill doesn't accidentally write files or call APIs it shouldn't. Catches problems in your own shell before anyone else runs the skill.

When to use it

Before adding a new skill to your .claude/ directory. Before sharing a skill publicly, where a silent misfire affects more than one machine.

Setup

  1. 1

    Save the file as ~/.claude/skills/skill-tester.md.

  2. 2

    Needs a test-cases file (JSON or YAML) describing inputs and the expected behaviour for each.

  3. 3

    Invoke as /skill-tester <skill-name>; the skill reports pass/fail per case plus the first failure details.

  4. 4

    Dry-runs simulate but don't guarantee — still ship to a test environment before production.

Example

You: /skill-tester pivot-sweep
Claude: 3 test cases. 2 pass ✓. 1 fails: tier-boundary case — treats a prompts/*.js file as Tier B instead of Tier A. Review regex.

Source & attribution

Author
Bryan Collins
Licence
MIT
Type
Original

Original pattern published under MIT — attribution preserved by convention, not licence requirement.

Caveats

Dry-run simulates but doesn't guarantee — ship to a test environment before production.

New skills, explained plainly.

One short email when a new Claude Code skill is worth installing — and when to skip the hype. No spam.

56 skills across 10 categories, all licence-verified.