Single tool 'db2toon.extract_schema' has a well-structured JSON Schema with proper type definitions, enum constraints, and nested object validation. Description is comprehensive (280+ chars) and explains WHAT, WHEN, and prerequisites. Parameters are well-documented with descriptions and constraints (minLength, minimum, maximum, default values). However, output schema is not documented in the visible code, the tool description does not specify what fields the TOON format response contains, forcing LLMs to infer structure. Error handling guidance is absent. Tool naming is clear but could be more action-oriented (extract_schema vs get_database_schema). No tool annotations (readOnlyHint) despite being explicitly read-only.

• Document the output schema explicitly. Add a 'Returns' section to the tool description specifying TOON format fields (e.g., 'Returns: {tables: [{name, columns: [{name, type, nullable}], indexes, constraints}], views: [...], metadata: {...}}'). This enables LLMs to extract and chain results.
• Add error recovery guidance. Extend description with: 'On connection errors, verify credentials and network access. On parse errors, check SQL dump syntax. On timeout, increase timeout parameter or reduce example_sample.'
• Add readOnlyHint annotation to the tool schema. Set 'readOnlyHint: true' to signal that this tool is safe to call without side effects, enabling client-side caching and safety optimizations.
• Clarify timeout format in parameter description. Change 'Maximum extraction duration as a Go duration' to 'Maximum extraction duration as a Go duration string (e.g., 30s, 5m, 1h). Defaults to 30s.'
• Consider renaming to 'get_database_schema' or 'describe_database_schema' to align with LLM verb-noun conventions and improve discoverability in tool lists.