calm-mcp exposes 2 tools with clear, domain-specific descriptions (194 - 250 chars each, within baseline 34 - 392 range). Both tools have documented return types and parameter descriptions. However, input schemas lack formal JSON Schema type declarations visible in the code, parameters are inferred from Python type hints only. The calm_critique tool is a stub returning templated responses, limiting practical utility. Error handling is minimal; no recovery guidance or validation feedback. Naming is clear (verb_noun pattern) but generic enough that tool selection is straightforward.

• Declare input schemas explicitly in FastMCP tool registration using JSON Schema format with type, description, enum, and default fields. Example: @mcp.tool() with explicit InputModel or schema dict.
• Complete calm_critique implementation to call Anthropic API (v1.1 goal) or remove the tool until functional. Stub tools reduce agent confidence and waste context.
• Add try-except blocks with actionable error messages. E.g., 'Invalid voice: got "xyz", must be one of: musk, jobs, ive, consensus' instead of silent failure.
• Document calm_ive_scan output schema in the docstring: 'Returns {"score": int (0-5), "issues": [{"line": int, "col": int, "type": str, "severity": int (1-3), "snippet": str, "suggestion": str}, ...]}'.
• Add input validation: enforce max text length (e.g., 100KB) and validate voice enum in calm_critique before processing. Return clear validation errors.