Single tool with incomplete schema and minimal descriptions. Tool name lacks action verb (estimate_semantic_entropy is acceptable but borderline). Input schema present but parameter descriptions are extremely terse (under 20 chars each). No output schema documented. No error handling guidance. No parameter constraints (enums, ranges, formats). Tool does one thing but definition quality is well below production baseline.

• Expand parameter descriptions to 50-100 chars each. Example: 'responses: Array of 2-20 sampled LLM completions (strings). Each should be a distinct generation from the same prompt to measure semantic consistency.'
• Document the output schema in the tool description or as a separate schema field. Include all returned fields: semantic_entropy (float, 0-log(N)), cluster_count (int), is_hallucination (bool), cluster_sizes (array), dominant_cluster_fraction (float 0-1), confidence_score (float 0-1).
• Add input validation constraints: minItems=2 for responses, maxItems=50 (or document the limit). If log_likelihoods provided, require length to match responses length.
• Enhance tool description with use case: 'Call this to detect hallucinations by measuring semantic entropy across N stochastic LLM completions. Low entropy (< 0.65 nats) indicates factual consensus; high entropy suggests contradictory outputs (hallucination risk).'
• Add error handling guidance in description: 'Returns error if responses is empty or log_likelihoods length does not match responses length. Ensure at least 2 responses for meaningful entropy calculation.'