audit-agent

Auditing Claude Code agents involves validating definition files against 12 anti-patterns and frontmatter rules to identify missing fields, tool-allowlist issues, and model mismatches, returning a structured report for remediation.

Common anti-patterns in agent definitions include missing frontmatter fields like name or description, incorrect model hints, and tool permission issues. These quality defects are detected by checking the agent body against established best-practice standards.

Yes, you can use an agent audit tool during code reviews or CI checks to validate agent definitions. It processes agents/<name>.md files and returns a structured compliance report highlighting severity levels for actionable remediation.

Validate frontmatter fields and tool allowlists by running an audit against your agents/<name>.md files. This checks for required metadata like name, description, and model hints, while detecting tool-allowlist issues and other permission defects in the agent body.

The best way to detect model mismatches and missing fields is to audit agent definitions against predefined frontmatter rules and anti-patterns. This validates references and returns a structured report highlighting severity levels for actionable remediation.