CHIM MCP server has 16 read-only tools with basic schemas and descriptions, but significant gaps prevent production use. All tools lack output schema documentation, LLMs cannot predict response structure. Most descriptions are adequate (50-150 chars) but lack context on WHEN to use each tool vs. similar ones (e.g., query_eventlog vs. get_speech_log). Parameters are typed but many lack descriptions (e.g., search_oghma has empty properties object). No error handling guidance, no recovery hints. Tools like run_query and read_file expose dangerous capabilities (arbitrary SQL, filesystem access) with minimal validation hints. Naming is clear (verb_noun pattern) but composition is weak, multiple overlapping tools (get_memories, get_memory_summaries, query_eventlog, get_diaries) suggest the agent must reason about subtle distinctions without explicit guidance.

• Document output schemas for all 16 tools. Include field names, types, and descriptions. Example: query_eventlog should return {events: [{id, timestamp, speaker, type, content}], total_count, has_more}.
• Expand search_oghma input schema. Add parameters like 'query' (string, required), 'limit' (number, default 50), 'category' (enum or string). Currently the tool is unusable because LLM cannot determine what to pass.
• Add error handling guidance to run_query and read_file. Example: 'If SQL syntax error occurs, check table names via list_connectors. If path not found, use list_files to discover valid paths.'
• Clarify tool distinctions in descriptions. Add 'Use query_eventlog for dialogue/events, get_speech_log for speech records, get_diaries for diary entries, get_memories for NPC memory records.' to each description.
• Add pagination guidance to list_* and search_* tools. Document that results are capped at limit (default 50) and explain how to use offset/limit for pagination. Include total_count in responses.