This server exhibits significant definition quality gaps typical of early-stage academic/research tools. Of 8 tools, only 2-3 have adequate parameter descriptions. Most tools lack detailed output schemas, making it difficult for LLMs to plan chained calls. Parameter types are present in JSON schemas but descriptions are sparse or missing entirely. Tools like `estimate_chemical_dosing_tool` and `compute_bulk_composites` accept complex nested objects (`custom_params`, `adm1_state`) with minimal guidance on required subfields. Error handling is present but generic (returns status/message without actionable recovery paths). The tool names are reasonably clear (verb_noun convention) but compositions suggest multiple concerns bundled into single tools (e.g., `elicit_basis_of_design` both collects AND validates AND stores state). Output documentation is almost entirely absent, no tool explicitly documents what fields the response will contain, forcing LLMs to infer structure.

• Document output schemas for all 8 tools. For each tool, add a Returns section specifying field names, types, units, and example values. E.g., get_design_state should document: Returns {basis_of_design: {feed_flow_m3d: float, cod_mg_l: float, ...}, adm1_state: {...}, ...}.
• Replace generic 'object' parameters with formally defined nested schemas. For estimate_chemical_dosing_tool.custom_params, define: {sulfide_mg_L: {type: number, minimum: 0, maximum: 1000, description: 'Sulfide concentration in mg/L'}, ...}.
• Refactor tools 5, 7, 8 (validate_adm1_state, compute_bulk_composites, check_strong_ion_balance) to return structured results instead of CLI instructions. E.g., validate_adm1_state should return {status: 'pass'|'fail', violations: [{component: 'S2_minus', expected: 150, actual: 145, error_percent: 3.3}], imbalance_meq_L: 0.2, message: '...'}.
• Add explicit ADM1 state schema documentation. Create a reference table showing all 30 component names, types, and units (e.g., X_su: float in kg COD/m³, S_su: float in kmol/m³). Link from all tools that accept adm1_state parameter.
• Upgrade reset_design description: 'DESTRUCTIVE: Irreversibly resets all design state (basis_of_design, adm1_state, results). All unsaved data is lost. Cannot be undone. Requires explicit confirmation.'
• Add unit and range constraints to numeric parameters. E.g., tolerance in validate_adm1_state: 'Relative tolerance for validation (0.0 - 1.0, default 0.1). E.g., 0.1 = 10% acceptable error.'
• Implement error responses following recovery-guide pattern. Template: '{"status": "error", "error_code": "invalid_parameter_group", "message": "Invalid parameter_group: solids_extra. Must be one of: essential, solids, nutrients, alkalinity, all.", "recovery_suggestion": "Try calling again with parameter_group=essential for minimum required parameters."}'