Three read-only tools with basic descriptions and partial schemas. Tool names follow verb_noun convention (kg_search, kg_get_node, kg_list_proposals). Descriptions exist but are brief (40-60 chars) and lack context on when to use each tool or what downstream actions they enable. Input schemas are present with type declarations, but parameter descriptions are minimal (10-20 chars). Output is JSON strings rather than structured objects. No error handling guidance, no pagination support despite search returning variable results, and no indication of result limits. The server is STDIO-only, which is a hard transport cap.

• Expand tool descriptions to 100-150 chars. For kg_search: 'Search the knowledge graph by keyword across node titles, descriptions, and markdown bodies. Use this to discover relevant nodes before fetching details with kg_get_node. Returns up to 5 results by default; use limit to adjust.'
• Convert parameter descriptions to formal constraints. For kg_search.role, declare as enum: {"type": "string", "enum": ["builder", "domain"], "description": "Filter by node role: builder (framework) or domain (subject-matter)."} instead of prose examples.
• Return structured objects instead of JSON strings. Change kg_search to return {"query": str, "count": int, "results": [{"id": str, "title": str, "description": str, "role": str, "track": str}]} as a dict, not json.dumps().
• Document result limits and pagination. Add to kg_search description: 'Results are limited to 50 nodes maximum. Use limit parameter (1-50) to control page size. No cursor pagination yet; consider offset-based paging for large result sets.'
• Add error recovery guidance. For kg_get_node, return: {"error": "node not found: <id>", "recovery": "Try kg_search() with keywords from the node title or description to find the correct node_id."}

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry