Server has 4 tools with basic schemas and descriptions, but significant gaps prevent production use. All tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization), and input schemas are present with typed parameters. However, parameter descriptions are minimal or absent, output schemas are undocumented, and error handling provides no recovery guidance. Tool names follow verb_noun convention (list-, get-, update-, create-) but descriptions lack context on WHEN to use each tool or what they return. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. The server lacks pagination support, result limits, and field-level documentation needed for agent composition.

• Expand tool descriptions to 50-200 characters. Example: 'list-sources: List all source files in a repository directory. Returns file names, types, and sizes. Call this first to discover available sources before fetching individual files with get-source.'
• Add parameter descriptions for all inputs. Example: 'org: GitHub organization name (e.g. mycompany). repo: Repository name within the org (e.g. docs). path: Directory path relative to repo root (e.g. src/components). Use forward slashes; leading/trailing slashes optional.'
• Document output schemas for each tool. Example for list-sources: 'Returns {files: [{name: string, type: "file"|"dir", size: number, path: string}], total: number, nextCursor?: string}'. For get-source: 'Returns {content: string, path: string, language: string, size: number}'.
• Add tool annotations to schema registration. Mark list-sources and get-source with readOnlyHint=true. Mark update-document and create-page with destructiveHint=true. This enables agents to reason about retry safety.
• Implement pagination for list-sources. Add limit (default 20, max 100) and cursor parameters. Return nextCursor in response. Document: 'Large repos may have hundreds of files; pagination prevents context window exhaustion.'