Two tools with basic descriptions and partial schemas. qdrant-find and qdrant-store both have short descriptions (18-27 chars, well below the 50-200 char LLM-optimized baseline) and lack output schema documentation. Input schemas are present but incomplete: query_filter is typed as 'object' with no structure guidance, and metadata similarly lacks constraints. No error handling guidance, no parameter validation rules stated, no recovery paths documented. Tool names follow verb_noun pattern (find, store) but descriptions are too terse to guide LLM selection. Missing: output schemas, error classification, actionable error messages, parameter constraints (e.g., max query length, metadata size limits).

• Expand tool descriptions to 80-150 chars. Example: 'qdrant-find: Search for stored memories by semantic similarity. Returns matching entries with content and metadata. Use this to retrieve context before responding to user queries. Requires a valid collection_name.'
• Document output schemas explicitly. For qdrant-find, specify: returns array of {id, content, metadata, score}. For qdrant-store, specify: returns {success: bool, entry_id: string, message: string}.
• Add parameter constraints to descriptions. For query: 'Text to search for (1-1000 chars, required)'. For metadata: 'JSON object with arbitrary keys (max 10KB, optional)'. For query_filter: 'Qdrant filter object (see Qdrant docs for syntax; optional)'.
• Add pagination to qdrant-find: include limit (1-100, default 20) and offset (default 0) parameters. Return total_count in response so LLM knows if more results exist.
• Document error cases and recovery paths. Example: 'If collection_name not found, call list_collections() first. If query is too long, truncate to 1000 chars and retry.'

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

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