Server has 11 tools with complete JSON schemas and descriptions. Naming follows verb_noun convention (sync_*, list_*, get_*, search_*, save_*, recommend_*, export_*). Descriptions are present (avg ~80 chars) but lack depth on WHEN to use each tool and dependencies. No parameter-level documentation visible in schema. Output schemas not documented. Error handling not visible in source. Risk classification (WRITE vs READ_ONLY) is present but not surfaced to LLM.

• Add parameter descriptions to all tools. For each param, document: what it controls, valid range/enum, and format. E.g., 'limit (integer, 1-100): Maximum number of results to return. Defaults to 20.'
• Document output schemas for all tools. Specify which fields are returned, their types, and which IDs can be chained to other tools. E.g., 'Returns array of {full_name: string, language: string, stars: integer, analysis: {category, tags}}'.
• Enhance tool descriptions with WHEN-to-use guidance. E.g., 'Call sync_stars first to fetch latest from GitHub, then list_star_changes to see what changed since last sync.'
• Add error handling guidance. Document what happens on GitHub API errors, rate limits, missing repos, and invalid queries. Provide recovery hints: 'If sync fails due to rate limit, wait 1 hour or check GITHUB_TOKEN.'
• Clarify pagination behavior. Document whether limit is enforced, whether results are sorted, and whether a total_count or next_cursor is returned for continuation.