This server exhibits major gaps across naming, descriptions, schemas, and error handling. Tool names lack consistency and clarity. Descriptions are present but minimal and lack LLM-optimization guidance. Input schemas are visible but lack proper type constraints and detailed parameter documentation. Error handling is absent, functions return None or generic error strings without recovery guidance. The server is not production-ready.

• Rename tools to follow verb_noun pattern with clear distinctions: 'get_pr_metadata' (PR title, author, state) and 'get_pr_file_changes' (file diffs and stats). Avoid 'fetch_pr' and 'fetch_pr_changes' overlap.
• Expand all descriptions to 80 - 200 chars. Include: WHAT it does, WHEN to use it vs similar tools, WHAT it returns. E.g., 'Fetch metadata and description for a pull request. Use this to get PR title, author, and state. Use get_pr_file_changes to see diffs. Returns pr_info object with title, author, created_at, updated_at, state.'
• Rename 'analyze_pr_with_openai' to 'generate_pr_analysis' and remove provider name from tool interface. Move OpenAI selection into config or environment.
• Add comprehensive input schemas with parameter descriptions and constraints: pr_number (integer, 1 - 1000000), repo_owner (string, 1 - 39 chars, alphanumeric + hyphen), repo_name (string, 1 - 255 chars), pr_data (object with required fields title, author, changes).
• Document output schemas for all tools. E.g., fetch_pr_changes returns: { title (string), author (string), changes (array of { filename (string), status (string: added|modified|removed), additions (integer), deletions (integer), patch (string) }) }.
• Implement structured error handling: return { status: 'error', code: '<code>', message: '<actionable message>', recovery: '<next step>' }. E.g., 'status: error, code: PR_NOT_FOUND, message: PR #999 not found in USERNAME/REPO_NAME, recovery: Verify repo owner and PR number using list_pull_requests.'