Single tool with incomplete definition quality. The tool 'summarise_pbip' has a basic description and input schema, but lacks critical elements: no output schema documentation, minimal parameter description, no error handling guidance, and no enum/constraint definitions. The tool performs a file-system read operation but provides no validation, no pagination support for potentially large results, and no recovery guidance for edge cases (missing project_path, permission errors, malformed JSON files). Tool name is acceptable but description is generic and under-optimized for LLM invocation.

• Add explicit output schema documentation: {'type': 'object', 'properties': {'PBIP_File': {...}, 'Report_Folder': {...}, 'SemanticModel_Folder': {...}, 'GitIgnore': {...}}, 'required': [...]}. Clarify which fields may be null or missing.
• Expand 'project_path' parameter description: 'Absolute or relative path to the PBIP project root directory. Must contain .PBIP or .PBIR file. Returns error if path does not exist or is not readable.'
• Add input validation: reject paths containing '..', validate that project_path is a directory, check read permissions before attempting file operations. Return structured error: {'error': 'Path not found', 'path': <sanitized_path>, 'recovery': 'Verify the project path exists and is readable.'}
• Implement result size limits: cap file content to first 10KB per file, add 'truncated' flag if content exceeds limit, document limit in description.
• Add error handling with recovery guidance: 'If JSON parsing fails, returns {"error": "Malformed JSON", "file": <filename>, "recovery": "Verify JSON syntax in project files."}'
• Document the hardcoded folder structure in the description, or make folder names configurable as optional parameters.
• Add a 'fields' parameter with enum constraint (e.g., ['pbip_metadata', 'reports', 'semantic_model', 'gitignore']) to allow filtering returned sections and reduce payload size.