Review and validate Claude Code agents for frontmatter correctness, description quality, system prompt structure, tool scope appropriateness, and model tier justification. Use when reviewing agents after creation, auditing existing agents for compliance, validating agent changes before committing, checking triggering examples, or auditing all agents in a plugin for consistency.
Reviews Claude Code agent definition files for structure, quality, and correctness.
| Step | Focus |
|---|---|
| 1. Location | Correct directory, naming conventions |
| 2. Frontmatter | Required fields, format, value constraints |
| 3. Description | Trigger quality, examples, specificity |
| 4. System prompt | Role, structure, output format, tone |
| 5. Tool scope | Least privilege, read-only vs write alignment |
| 6. Model tier | Tier matches agent complexity |
| 7. Issue identification |
| Severity classification |
| 8. Recommendations | Actionable improvement items |
Read-only — never modify reviewed files. Produce a YAML verdict only.
See references/process.md for the full review methodology: 8-step process,
spec compliance checks, quality review dimensions, verdict format, and common findings.