Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. 'measure' and 'diff' both have descriptions (good baseline), but parameter descriptions are minimal and output schemas are not documented. No error handling guidance, no recovery patterns, and no tool annotations. Naming is acceptable (verb-noun style) but descriptions lack context on when to use each tool vs. the other. Parameter 'runs' lacks range constraints; 'mode' enum is present but sparse. No evidence of idempotency guarantees or composition guidance.

• Document output schemas for both tools. E.g., 'measure returns {lcp_ms: number, inp_ms: number, cls_score: number, timestamp: string, raw_trace?: string}'. This enables LLMs to extract and chain results.
• Add range constraints to 'runs' parameter: 'Number of measurement runs (1-100; default 3). More runs increase accuracy but take longer.'
• Expand parameter descriptions with format and dependency hints. E.g., 'url: Full HTTP(S) URL to measure (e.g., https://example.com). Must be publicly accessible or on localhost.'
• Add error handling guidance in tool descriptions. E.g., 'If measurement fails due to network timeout, the tool returns {error: string, retryable: boolean}. Retryable errors can be retried after 30s.'
• Clarify when to use 'diff' vs 'measure': 'Use measure() to get metrics for a single URL. Use diff() to compare performance between two URLs (e.g., before/after a deployment).' This guides LLM tool selection.