Policy-governed multi-server aggregation, sandboxing, and projection for Model Context Protocol servers with encrypted vault management and audit logging
This server exposes 9 tools with significant definition gaps. Most tools lack parameter descriptions entirely (e.g., gitee__list_pull_requests, gitee__get_commit_diff have empty input schemas with no property descriptions). Tool descriptions are present but minimal (10-50 chars), below the 50-200 char baseline for LLM optimization. No output schemas are documented. Error handling is absent, no recovery guidance, categorization, or actionable error messages. The server acts as a forward proxy aggregating multiple backends (Gitee, workspace FS, Context7), but tool composition is unclear: it's not evident which tools chain together or what IDs they exchange. Security concerns: no visible permission gating, audit logging exists but tool-level scope declarations are missing.
Fetch up-to-date documentation content from Context7 index
Resolve package or library name to Context7 ID (e.g. /facebook/react, /vercel/next.js)
Create a new issue on Gitee repo
Get commit diff and ast tree
List open pull requests on Gitee repo
Verify end-to-end responsiveness of the server
Get real-time operational status of the server
Empty input schemas with no property descriptions. gitee__list_pull_requests, gitee__get_commit_diff, ping, and status have type:object with no properties defined. LLMs cannot infer what parameters to pass.
No output schemas documented. Callers cannot predict response structure, forcing LLMs to guess field names for downstream tool chaining. Breaks the tool-chain pattern.
Minimal tool descriptions (10-50 chars). Below 50-char baseline for LLM optimization. E.g., 'Verify end-to-end responsiveness of the server' (ping) lacks context on when to call it vs status. Descriptions should be 50-200 chars with WHAT, WHEN, and prerequisites.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
List files in scoped workspace directory
Read file from scoped workspace
No error handling guidance. No recovery hints, error categorization, or actionable messages. If a tool fails, LLMs have no direction on retry, user input, or fallback.
No tool-level permission declarations or scope hints. Audit logging exists server-wide, but individual tools lack scope metadata (e.g., 'read:gitee', 'write:gitee'). Prevents least-privilege agent configuration.