One command. Your codebase, AI-ready. MCP configuration manager for Claude Desktop integration.
Three tools with basic verb-noun naming (read_file, list_directory, search_content) and short descriptions (10-60 chars). All have input schemas with type and description fields. However, descriptions are minimal and lack context on when/why to use each tool. No output schemas documented. No error handling guidance. Parameter descriptions are terse. Tools are read-only and safe, but lack the depth expected for production use.
List files and directories in a directory.
Read and return the contents of a file.
Recursively search for files containing query, returning matching paths and lines.
Descriptions are under 20 characters and lack context. 'Read and return the contents of a file' (47 chars) is minimal; should explain when to use vs list_directory, what formats are supported, and size limits.
No output schemas documented. Tools return string or list[str], but LLMs cannot infer structure. search_content returns 'matching paths and lines' but format is undocumented (is it 'file:line: content' or JSON?).
Error handling returns plain strings ('Error: path is outside BASE_DIR') with no guidance on recovery. LLMs cannot distinguish retryable from fatal errors or know what to do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
Parameter 'path' description is generic ('File path relative to BASE_DIR'). Should specify format, constraints (max length, allowed characters), and examples of valid paths.
search_content lacks pagination. No limit parameter or result count cap documented. Recursive search on large codebases could return thousands of lines, exhausting context window.