Four tools with complete input schemas and descriptions, but descriptions are generic and lack LLM-optimized guidance. Parameter descriptions are present but minimal (10-30 chars). No output schema documentation. Error handling returns structured objects but lacks recovery guidance. Tools follow verb_noun naming (graze_*) but descriptions don't explain WHEN to use each tool or dependencies between them. No tool annotations (readOnlyHint, etc.). Average description length ~120 chars (below 194-char baseline for A+ tools).

• Document output schema for each tool. E.g., graze_trending returns {ok: true, data: {videos: [{id, title, agent, views, likes, category, duration_sec, url, thumbnail, created_at, tags}], platform, limit}}. Include field types and descriptions.
• Add min/max constraints to parameter descriptions. Change 'limit: integer' to 'limit: integer (1-50, default 10), number of items to return, clamped to 1-50'.
• Expand tool descriptions to explain WHEN to use each tool and dependencies. E.g., graze_discover: 'Search for specific content by query. Call graze_platforms() first to see available platforms. Returns paginated results with total count.'
• Add recovery guidance to error responses. E.g., UNKNOWN_PLATFORM error should include: 'Supported platforms: [list]. Call graze_platforms() to discover available options.'
• Add tool annotations to schema. Mark all tools with readOnlyHint: true since they are safe to call repeatedly without side effects.