Secure MCP server with OAuth 2.0 for ChatGPT compatible with ChatGPT Developer Mode (JSON-RPC over HTTP) for reading local files
Server has 4 tools with basic schemas and minimal descriptions. All tools have input schemas with type definitions, but descriptions are extremely brief (10-30 chars), falling well below the 50-200 char baseline for LLM-optimized tools. No output schemas documented. Parameter descriptions are generic ('Directory path relative to BASE_DIR'). No error handling guidance, no recovery hints, no parameter constraints (enums, ranges). Destructive operations (write_file, delete_file) lack confirmation patterns or dry-run support. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Security concerns: path parameter accepts relative paths with no visible traversal validation in source.
Delete a file
List files in a directory
Read file contents
Write content to a file
Descriptions critically short (10-30 chars). LLMs cannot determine when/why to select tools. Baseline is 50-200 chars.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., file list structure, read content format).
Destructive operations (write_file, delete_file) lack confirmation/dry-run pattern. Agents can cause irreversible data loss without safeguards.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. LLMs cannot reason about operation safety.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | <=2025-11-25 | v2 |
Path traversal validation not visible in source. 'path' parameter accepts relative paths with no documented constraints or sanitization.