Server has 9 tools with complete schemas and descriptions. Naming follows verb_noun convention (list_, fetch_, generate_, post_, review_, open_). Descriptions are present (avg ~80 chars) but lack LLM-optimized guidance on WHEN to use each tool and dependencies between them. Parameters are typed with descriptions, but lack constraints (enums, ranges, patterns). Error handling is minimal, no recovery guidance or actionable error messages visible in tool definitions. Output schemas are not documented. Tools are well-composed (single responsibility) but lack pagination support for list_open_prs. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels.

• Add tool annotations: mark post_pr_review and review_pr (post=true) with destructiveHint=true; mark all others with readOnlyHint=true. This enables LLMs to reason about safety.
• Document output schemas for all tools. E.g., list_open_prs returns [{pr_number, title, author, state, url}, ...]; fetch_pr_metadata returns {pr_number, title, body, author, state, created_at, updated_at, ...}.
• Enhance descriptions with dependency hints: 'Call fetch_pr_diff() first to load the diff, then call generate_pr_review() to analyze it.' This guides multi-step planning.
• Add enum constraints to provider parameter: valid values are ['claude', 'antigravity', 'codex'] (or actual list). Include in schema and description.
• Add pagination to list_open_prs: accept offset/limit parameters, return total_count and has_more fields. Document the default limit (20) and max limit in description.