ReviewGenie has 5 tools with basic schemas and descriptions, but significant gaps prevent production use. Tool names lack clear action verbs (ping, ingest_pr, analyze_pr are vague about what they return vs. modify). Descriptions exist but are brief (avg ~80 chars) and lack WHEN/WHY guidance. Parameters have types but most lack descriptions (e.g., 'style' param in inline_comments has a description, but 'framework' in generate_tests does not). No output schemas documented. Error handling is minimal, exceptions are caught but not classified or recovery-guided. The inline_comments tool accepts a dry_run parameter, which is good, but generate_tests lacks idempotency guarantees and could create duplicate PRs on retry.

• Rename tools to action verbs: 'ping' → 'get_pr_metadata', 'ingest_pr' → 'fetch_pr_diff', 'analyze_pr' → 'analyze_pr_quality', 'inline_comments' → 'post_review_comments', 'generate_tests' → 'create_test_pr'.
• Document output schemas for all tools. E.g., ping returns {title: str, author: str, state: str}; analyze_pr returns {smells: [{type: str, line: int, severity: str}], risk_score: float}.
• Add descriptions to all parameters. For 'pr_url', specify format (e.g., 'GitHub PR URL in format https://github.com/owner/repo/pull/123'). For 'framework', list supported values ('pytest' only currently).
• Implement error classification and recovery guidance. E.g., 'Comment failed: rate limit exceeded. Retry in 60 seconds.' or 'Comment failed: PR is closed. Cannot post reviews to closed PRs.'
• Add idempotency to generate_tests. Check if a test PR already exists for this PR number before creating a new branch. Return the existing PR URL if found.