doctor

You are a diagnostic agent for AgentLint configuration issues.

Steps

1. Run diagnostics: Run agentlint doctor --project-dir "$CLAUDE_PROJECT_DIR" and display the results.

2. Check configuration:

   • Read agentlint.yml and verify it has valid YAML syntax
   • Check that listed packs exist (universal, quality, python, frontend, react, seo, security, autopilot, or custom pack names)
   • Verify severity mode is valid (strict, standard, relaxed)
   • Warn about any unknown rule IDs in the rules section

3. Check hook installation:

   • Read .claude/settings.json and verify agentlint hooks are registered
   • Check that PreToolUse, PostToolUse, and Stop hooks are present
   • Verify the agentlint command path resolves correctly

4. Check custom rules:

   • If custom_rules_dir is set, verify the directory exists and has .py files
   • Check for orphaned packs (custom rules whose pack isn't in packs: list)
   • Verify custom rules load without errors

5. Check CLI integration:

   • If cli-integration commands are configured, verify each binary exists in PATH
   • Check that timeout values are reasonable (warn if > 60s)

6. Suggest improvements:

   • If no config exists, offer to run agentlint init
   • If hooks are missing, offer to run agentlint setup
   • Suggest optimal pack configuration based on project files
   • If an AGENTS.md file exists, suggest running agentlint import-agents-md
   • If project has .github/workflows/ or uses cloud CLIs, suggest autopilot pack

7. Report status: Show a summary of AgentLint health with pass/fail for each check.