| name | evaluate-skill |
| description | Evaluate a skill by running test cases and grading results. Use when testing whether a skill produces correct guidance, validating improvements, or benchmarking before release. |
| args | <plugin/skill-name> [--create-evals] [--runs N] [--baseline] |
| allowed-tools | Task, Read, Write, Edit, Glob, Grep, Bash(bash *), TodoWrite |
| argument-hint | git-plugin/git-commit [--create-evals] [--runs 3] [--baseline] |
| agent | general-purpose |
| context | fork |
| created | "2026-03-04T00:00:00.000Z" |
| modified | "2026-07-18T00:00:00.000Z" |
| compatibility | claude-code |
| reviewed | "2026-03-04T00:00:00.000Z" |
/evaluate:skill
Evaluate a skill's effectiveness by running behavioral test cases and grading the results against assertions.
When to Use This Skill
| Use this skill when... | Use alternative when... |
|---|
| Want to test if a skill produces correct results | Need structural validation -> scripts/plugin-compliance-check.sh |
| Validating skill improvements before merging | Want to file feedback about a session -> /feedback:session |
| Benchmarking a skill against a baseline | Need to check skill freshness -> /health:audit |
| Creating eval cases for a new skill | Want to review code quality -> /code-review |
Context
- Available skills: !
find . -path '*/skills/*' -name 'SKILL.md' -not -path '*/.claude/worktrees/*'
Parameters
Parse these from $ARGUMENTS:
| Parameter | Default | Description |
|---|
<plugin/skill-name> | required | Path as plugin-name/skill-name |
--create-evals | false | Generate eval cases if none exist |
--runs N | 1 | Number of runs per eval case |
--baseline | false | Also run without skill for comparison |
Execution
Step 1: Resolve skill path
Parse $ARGUMENTS to extract <plugin-name> and <skill-name>. The skill file lives at:
<plugin-name>/skills/<skill-name>/SKILL.md
Read the SKILL.md to confirm it exists and understand what the skill does.
Step 2: Run structural pre-check
Run the compliance check to confirm the skill passes basic structural validation:
bash scripts/plugin-compliance-check.sh <plugin-name>
If structural issues are found, report them and stop. Behavioral evaluation on a structurally broken skill is wasted effort.
