audit-agents-skills — Claude Code command: agent quality audit
audit-agents-skills — a command for Claude Code and Codex that audits agent, skill, and command quality in a project. Scans files, scores them against criteria (identity, prompt, validation, design), and grades production readiness.
CC-BY-SA-4.0 ★ 6,529 updated 2026-10-22
.claude/commands/audit-agents-skills.mdWhat audit-agents-skills does
The command runs in three phases: discovers all agents, skills, and commands in .claude/ directories, scores each file against weighted criteria (agents and skills out of 32 points, commands out of 20), and outputs readiness grades. Use --fix to generate improvement suggestions; --verbose shows details for all criteria, not just failures.
For agents, it checks name and description (with usage context), whether a model is specified, tool restrictions, role definition in the prompt, output format, and edge case documentation. For skills, it validates YAML structure, enforces name format [a-z0-9-]+, checks for methodology and usage examples, and detects hardcoded paths or secrets. Commands are validated for name, description, arguments, and call examples.
The command scans project directories (.claude/agents/, .claude/skills/, .claude/commands/ and examples/ folders), classifies files by type, and outputs a report with counts and scores for each item.
What it helps with
- audit agent quality before deploying to CI/CD
- find duplicated functionality across agents and skills
- verify agents don't contain hardcoded values or secrets
- assess documentation completeness (examples, edge cases, error handling)
- identify agents missing role definition or output format
Example prompts for the agent
- Audit agent quality in the current project:
/audit-agents-skills - Run audit with fix suggestions:
/audit-agents-skills --fix - Get detailed report for all criteria:
/audit-agents-skills --verbose
How to install audit-agents-skills
Manually: put the files in place
.claude/commands/audit-agents-skills.mdCall it with /audit-agents-skills. For all projects use ~/.claude/commands/
Manually: put the files in place
~/.codex/prompts/audit-agents-skills.mdCall it with /prompts:audit-agents-skills. Codex marks this format deprecated in favour of skills
Manually: put the files in place
.github/prompts/audit-agents-skills.prompt.mdManually: put the files in place
.gemini/commands/audit-agents-skills.tomlGemini stores commands as TOML: the prompt text goes into the prompt field
Questions about audit-agents-skills
How do I install audit-agents-skills in Claude Code?
Copy the command file to .claude/commands/: create the directory if needed and place audit-agents-skills.md inside. The command becomes available immediately in Claude Code.
What does audit-agents-skills do?
The command scans all agents, skills, and commands in your project, scores each one against criteria (name, description, prompt quality, examples, security), and outputs readiness grades. Use --fix to see what needs improvement; --verbose shows details for every check.
Which criteria does the command check?
For agents: name and description with usage context, specified model, tool restrictions, role definition, output format, edge cases, file size. For skills: valid YAML structure, name format [a-z0-9-]+, documented methodology, usage examples, no hardcoded paths or secrets. For commands: name, description, arguments, examples.