Five tools with clear verb-noun naming (search_paper, find_repository, list_repository_files, search_code, read_file). All have descriptions (avg ~150 chars, within baseline 194). However, input schemas are visible but lack critical structure: no type declarations beyond 'string', no enums for constrained inputs (e.g., branch names), no output schemas documented. Parameter descriptions exist but are minimal (avg ~40 chars vs baseline 72). No error handling guidance, no pagination support for list_repository_files despite returning potentially large trees. Tool descriptions state WHAT but not WHEN to use or prerequisites. Missing idempotency hints, rate-limit guidance, and recovery paths.

• Document output schemas for all tools. E.g., search_paper returns {context: string, page_numbers: int[], chunk_ids: string[]}; find_repository returns {repositories: [{name, url, stars, language, default_branch}], total_found: int}.
• Add pagination to list_repository_files: accept limit (1-500, default 100) and offset parameters; return {files: [...], total_count: int, truncated: bool}.
• Expand parameter descriptions to 60-100 chars. E.g., 'branch: Git branch name (defaults to repository default branch; use "main" or "master" for common cases).'
• Add error recovery guidance. E.g., 'Repository not found. Verify owner/repo format (e.g., torvalds/linux). Try search_code with fewer keywords if rate-limited.'
• Declare input constraints explicitly. E.g., extension enum: [py, js, ts, java, go, rs, cpp, c, rb, php]; keywords max 200 chars; path must start with / or be relative.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry