Four tools with clear names and reasonable descriptions, but significant gaps in schema completeness and parameter documentation. All tools are read-only lookup/analysis operations with enum-constrained inputs. Schemas are present but lack detailed output documentation. Descriptions are adequate (100-200 chars) but could better explain when/why to use each tool. No error handling guidance or recovery paths documented.

• Document output schemas for all tools. Specify the JSON structure returned by get_system_info (e.g. {neural_basis: string, behavioral_signature: string, ...}) so LLMs know what fields to extract.
• Add error recovery guidance: 'Unknown system: {system}. Valid systems are: SEEKING, RAGE, FEAR, PANIC_GRIEF, CARE, PLAY, LUST. Did you mean one of these?'
• Simplify match_patterns description: replace 'lexical prefilter' and 'token overlap' with plain language: 'Finds patterns whose recognition text shares words with your signals. Results are a shortlist to verify, not a diagnosis.'
• Refactor analyze_behavior to return structured JSON (steps array with step name, description, and guidance) instead of markdown. Enables LLMs to reason about each step programmatically.
• Add a 'confidence' or 'match_score' field to match_patterns results so LLMs can rank candidates and know when the prefilter found nothing useful (below CONFIDENCE_FLOOR).