Single tool with minimal definition quality. Tool name 'decompose_uncertainty' lacks action verb prefix (should be 'analyze_uncertainty' or 'compute_uncertainty'). Description is adequate (89 chars, within 10-1024 range) but lacks WHEN to use guidance. Input schema present but parameter description is vague ('List of probability distributions') without format/constraint details. No output schema documented, LLMs cannot plan downstream calls. No error handling guidance. Tool does one thing well but definition falls short of production standards.

• Rename tool to 'analyze_uncertainty' or 'compute_uncertainty_decomposition' to start with action verb and improve LLM intent recognition.
• Document output schema in tool definition: specify that response is object with fields {total_uncertainty: float, aleatoric_uncertainty: float, epistemic_uncertainty: float, epistemic_ratio: float, mean_distribution: {[class]: float}, recommendation: enum(FETCH_MORE_CONTEXT|ASK_CLARIFICATION|CONFIDENT_EXECUTION), explanation: string}.
• Expand parameter description to specify: 'Array of M probability distributions (M ≥ 1). Each element is object mapping class names to probabilities (must be non-negative, sum to ~1.0). All distributions should have same class keys. Example: [{"yes": 0.8, "no": 0.2}, {"yes": 0.6, "no": 0.4}]'.
• Add error handling guidance to description: 'Returns error if sample_predictions is empty, contains non-numeric probabilities, or has negative values. Recommendation thresholds: epistemic > 0.15 triggers FETCH_MORE_CONTEXT; aleatoric > 0.3 triggers ASK_CLARIFICATION.'
• Enhance tool description with use-case context: 'Use this to decompose ensemble model uncertainty into reducible (epistemic) and irreducible (aleatoric) components. Helps agents decide whether to gather more context, ask for clarification, or proceed confidently.'