Critical failures across all dimensions. Tool definitions exist only in test files (mcp_tool_annotations_test.go), not in actual tool registration code, this violates the 'must see explicit registration' rule, capping all tools at 50 max. Descriptions are present but extremely terse (10-50 chars), well below the 50-200 char LLM-optimized baseline. No input schemas visible for 43 of 45 tools, only 2 tools (witself.fact.confirm, witself.fact.reject) show partial schemas with 2 params each. No output schemas documented anywhere. Parameter descriptions are missing entirely for most tools. No error handling guidance, no recovery paths, no actionable error messages. Tool naming uses dot notation (witself.fact.set) rather than verb_noun convention (set_fact), making intent parsing harder for LLMs. No evidence of idempotency hints, permission gates, or security annotations beyond risk labels. This server is a skeleton definition with test coverage but no production-grade tool interface.

• Move tool definitions from test file to actual tool registration code (likely cmd/witself/server.go or similar). Ensure each tool is registered with explicit name, description, input schema, and output schema in the MCP server initialization.
• Add complete JSON Schema input definitions for all 45 tools. Minimum: type, description, and required fields. Example for witself.fact.set: {"type": "object", "properties": {"subject_id": {"type": "string", "description": "The subject ID for the fact"}, "content": {"type": "string", "description": "The fact content"}}, "required": ["subject_id", "content"]}.
• Expand all tool descriptions to 80-150 characters. Include: (1) what the tool does, (2) when to use it vs similar tools, (3) any prerequisites. Example: 'Set a fact for a subject. Use this to create or update a fact after confirming it via witself.fact.propose. Requires subject_id and fact content.'
• Rename tools to verb_noun convention: witself.fact.set → set_fact, witself.message.send → send_message, witself.fact.list → list_facts. This aligns with 90% of A+ tools and improves LLM intent parsing.
• Document output schemas and pagination for all list tools. Example: 'Returns {"facts": [{"id": string, "subject_id": string, "content": string}], "total": number, "next_cursor": string|null}'. Add limit (1-100, default 20) and cursor parameters to list tool inputs.