Selvedge provides 8 tools with reasonable naming discipline and mostly complete schemas. However, descriptions are sparse and inconsistent. Most tool descriptions (log_change, diff, blame, changeset, search, stale_decisions) are under 100 characters, below the production baseline of 194 chars and below the minimum of 10 - 1024 chars for adequate context. Parameter descriptions exist but are minimal. The server has decent structure but lacks the depth needed for production-grade tool discoverability. No output schema documentation visible. This is solidly C+ territory, better than median community servers but missing critical polish.

• Expand all tool descriptions to 100 - 250 characters. State WHAT the tool does, WHEN to use it vs. similar tools, and WHAT it returns. Example: 'diff' → 'Get the change history for a specific entity (file, function, DB column, etc.). Shows all events affecting that entity in chronological order. Use when you want to see how a single component evolved. Returns a list of change events with timestamps, authors, and change types.'
• Document output schemas for all tools. For each tool, specify the structure of the return value: field names, types, whether arrays are returned, and what pagination looks like. This enables LLMs to chain tools and extract relevant data.
• Clarify tool distinctions in descriptions. Explicitly state: 'diff' returns history of one entity; 'history' filters across all entities; 'changeset' groups related changes. Add examples like 'use diff to see all versions of a function; use history to find all changes made in a feature branch; use changeset to see all changes under a task ID.'
• Add input validation and error guidance. For tools that accept entity_path, describe the format (dot/slash notation) and document error cases: 'entity_path not found: available entities are [...]'. For search, document what happens with empty results.
• Implement pagination for history and search. Add limit (default 20, max 100) and offset/cursor parameters. Return total_count or next_cursor so LLMs can iterate without re-running unbounded queries.
• Add parameter descriptions for tools with minimal input hints. For search's 'query' param, explain: 'Free-text search across change messages, entity paths, authors, and change types. Supports: entity name (e.g., "login"), author email, change type ("add", "remove"), date range.'