Two tools with complete schemas and reasonable descriptions, but significant gaps in parameter documentation and error handling. Tool names follow verb_noun convention (generate_tech_stack, list_tech_stack). Descriptions are adequate (120-180 chars) but lack actionable context for LLM selection. Input schemas present with types, but parameter descriptions are minimal or missing. No output schema documentation. Error handling is basic (existence checks only). Missing guidance on when to use each tool vs the other, and no recovery hints for common failures.

• Add explicit output schema documentation to both tools. For generate_tech_stack: 'Returns text summary with file path and category breakdown.' For list_tech_stack: 'Returns JSON object with keys: project (string), total_tools (int), categories (object mapping category names to arrays of {name, desc} objects).'
• Enhance parameter descriptions with format and constraint details. Example: 'project_dir: Absolute path to project root directory. Must exist and be readable. Defaults to current working directory if omitted.'
• Add tool selection guidance in descriptions. Prefix generate_tech_stack with: 'Generates a visual HTML report. Use this when you need a shareable tech stack document.' Prefix list_tech_stack with: 'Returns structured JSON data. Use this for programmatic access or piping to other tools.'
• Implement actionable error messages with recovery hints. When directory not found, suggest checking the path. When no dependencies detected, list supported file types and suggest checking the directory contents.
• Document supported dependency file formats and tool categories in the tool descriptions or a separate discovery tool. LLMs benefit from knowing what 'languages, frameworks, databases, AI SDKs, testing, observability, security, and infrastructure tools' means in practice.