Single tool 'estimate_2sls' has a reasonable description (97 chars, within 10-1024 range) but critical schema and parameter documentation gaps. Input schema visible in code shows three array parameters with descriptions, but lacks type validation, min/max constraints, and error handling. No output schema documented. Tool does one thing (2SLS estimation) but lacks production-grade parameter validation, error recovery guidance, and security considerations. No tool annotations (readOnlyHint, etc.). Descriptions are present but minimal; parameters lack format/range constraints.

• Document output schema explicitly: {beta_iv: float, beta_ols: float, stage1_f_stat: float, stage1_r2: float, is_strong_instrument: bool}. Include interpretation guidance for each field.
• Add input validation: require arrays of length >= 3, numeric values (no NaN/Inf), and equal lengths across z, x, y. Return actionable error: 'Arrays must have equal length >= 3. Got z={len(z)}, x={len(x)}, y={len(y)}'.
• Enhance parameter descriptions with format/range: 'instrument_z: array of floats, length >= 3, representing exogenous instrument values. No NaN or Inf allowed.'
• Add error recovery guidance: 'If stage1_f_stat < 10, instrument may be weak. Consider alternative instruments or check for multicollinearity.'
• Rename tool to 'estimate_causal_effect_instrumental_variable' or add alias 'fit_2sls_model' to improve LLM discoverability and reduce domain jargon confusion.