Server has 10 tools with complete schemas and descriptions, but quality is uneven. Tool names are action-oriented (get_, list_, validate_, fit_, calculate_, analyze_, publish_) and mostly clear. Descriptions exist but many are verbose and prescriptive rather than LLM-optimized. Critical issue: descriptions contain procedural instructions ('Call get_analysis_choices first...') that belong in server instructions, not tool descriptions. Parameters have types and descriptions, but several lack constraints (enums, ranges). Output schemas are not documented. Error handling is minimal, no recovery guidance or categorization. Security concern: publish_analysis_results writes to user desktop without explicit confirmation. Overall, definitions are functional but not production-grade.

• Move procedural workflow instructions from tool descriptions to server.instructions. Tool descriptions should state WHAT the tool does, not HOW to use it in a workflow. E.g., fit_ecm description should be: 'Fit an equivalent circuit model to EIS data using least-squares optimization. Requires frequency, impedance arrays, and model name.' Remove 'Call get_analysis_choices first...'
• Document output schemas for all tools. Add a 'Returns' section to each tool docstring specifying field names, types, and meanings. E.g., fit_ecm returns {success: bool, parameters: {R0: float, ...}, rmse: float, ...}
• Add enum constraints to parameters: ecm_weighting ∈ [modulus, real, imaginary]; ecm_model should reference list_ecm_models output; drt_config should document required/optional keys.
• Implement error responses with recovery guidance. When validation fails, return {success: false, error: 'Invalid frequency array: must be sorted descending. Call validate_eis() to check your data.'}
• Add a confirm_publish_results tool or dry-run parameter to publish_analysis_results. Before writing to desktop, return {preview: {source_path, target_path, file_count}, confirm_required: true} and require explicit confirmation.