Four search tools with identical schemas and minimal differentiation. All tools have basic descriptions (40-50 chars) and proper JSON Schema with required/optional fields. However, descriptions lack context on WHEN to use each tool vs. alternatives, and parameter descriptions are generic. No output schema documented. Error handling returns plain text without recovery guidance. Tool names follow verb_noun pattern but lack semantic distinction, all four are search_* variants operating on the same API with identical parameters, creating LLM disambiguation burden.

• Consolidate four search tools into one parameterized tool: search_museum(search_term, category=['all'|'objects'|'people'|'documents'], limit=5, offset=0). Use an enum for category to make valid options explicit and reduce LLM confusion.
• Expand tool description to 100-150 chars explaining the difference: 'Search the Science Museum collection. Category filters: all (all records), objects (physical items), people (historical figures), documents (archives). Returns up to 5 results by default.'
• Document output schema in tool description: 'Returns {records: [{id, title, description, ...}], success: bool, message: string, number_of_records: int}'. Include sample field names so LLMs know what to extract.
• Add parameter constraints: 'limit must be 1-100 (default 5 to conserve tokens); offset must be >= 0; search_term must be 1-200 characters.'
• Improve error messages with recovery guidance: 'Search failed: API returned 429 (rate limited). Wait 60 seconds and retry. If error persists, check search_term format (1-200 chars, alphanumeric + spaces).' Categorize as retryable vs user-fixable.

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

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