Three tools with basic descriptions and partial schemas. review_diff and review_range have clear action verbs and reasonable descriptions (60-80 chars), but lack output schema documentation. Parameters are typed but minimally described. usage_summary has no parameters but lacks output schema. No error handling guidance, no parameter constraints (enums), no recovery paths. Naming is acceptable but descriptions fall short of LLM-optimization standards (target 50-200 chars). No tool annotations (readOnlyHint, etc.) despite all being read-only operations.

• Add output schema documentation to all tools. Example: review_diff returns {verdict: 'PASS'|'BLOCK', reason: string, tokens_used: number}. This enables LLMs to extract and chain results.
• Expand parameter descriptions to 50-100 chars each. For 'model', specify: 'Claude model ID (e.g., claude-haiku-4-5, claude-opus-4-1). Defaults to claude-haiku-4-5 for cost efficiency.' For 'diff', add: 'Unified git diff format (output of git diff origin/main..HEAD). Must not be empty.'
• Add readOnlyHint: true annotation to all three tools to signal they are safe to call without side effects.
• Implement error handling with recovery guidance. Example: if diff is empty, return 'No diff to review. Ensure the git range contains commits (e.g., git log origin/main..HEAD).' If model is invalid, return 'Unknown model. Valid options: claude-haiku-4-5, claude-opus-4-1, claude-sonnet-4-20250514.'
• Add enum constraint for 'model' parameter or validate server-side and return actionable error messages listing valid options.