eval-agent-md
Behavioral compliance testing for any CLAUDE.md or agent definition file. Auto-generates test scenarios from your rules, runs them via LLM-as-judge scoring, and reports a compliance score with per-rule pass/fail breakdown. Optionally improves failing rules via automated mutation loop. Use when: (1) testing whether your CLAUDE.md rules are actually followed, (2) evaluating an agent definition for role-boundary compliance, (3) dogfooding a skill's own SKILL.md. Triggers on: "eval", "compliance test", "test my CLAUDE.md", "check rules", "behavioral test", "/eval-agent-md". Do not trigger for: editing or writing CLAUDE.md rules, general code review, adding linting config, or any task that is not explicitly about testing behavioral compliance.
The canonical home for this skill is eval-agent-md in ravnhq/ai-toolkit
Quality
Validation
75%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 12 / 16 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
compatibility_field | 'compatibility' field must be a string, got object | Fail |
allowed_tools_field | 'allowed-tools' must be a string, got object | Fail |
metadata_field | 'metadata' should map string keys to string values | Warning |
referenced_paths_exist | Referenced path issues: 2 missing | Warning |
Total | 12 / 16 Failed | |
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.