pastoon defines 2 tools with clear, action-verb names (json-to-toon, toon-to-json) and reasonable descriptions (100-120 chars). Both tools have input schemas with typed parameters and enums for constrained values (delimiter, keyFolding). However, output schemas are not documented, LLMs cannot predict what fields to expect from responses. Parameter descriptions are present but minimal (10-30 chars), lacking context on when to use each option or what the output structure contains. No error handling guidance is visible in the tool definitions. The tools are simple and focused (single responsibility), but lack the depth needed for confident production use.

• Document output schemas for both tools. Specify: json-to-toon returns {toon: string, tokensSaved: number, originalSize: number} and toon-to-json returns {json: string, originalSize: number}.
• Expand parameter descriptions to 50-100 chars. For delimiter, add: 'Field separator in TOON output. Use comma for CSV-like data, tab for aligned columns, pipe for structured records.' For keyFolding, add: 'off = preserve all keys; safe = abbreviate common keys (e.g. id→i, name→n) to reduce tokens further.'
• Add error handling guidance to tool descriptions. E.g. 'If input is not valid JSON, returns error: "Invalid JSON: <details>". Try validating with a JSON parser first.'
• Include example input/output in descriptions (not as sample values, but as format hints). E.g. 'Converts {"name":"Alice","age":30} to TOON format with ~40% fewer tokens.'
• Document size limits and performance expectations. E.g. 'Optimized for JSON up to 1MB. Larger inputs may be slow. Returns error if input exceeds 10MB.'