model-context-builder

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists primarily of documentation and instructional content for evaluating MCP servers. It provides guidelines on creating robust, read-only questions for testing LLM tool usage.
  • [EXTERNAL_DOWNLOADS]: Mentions an evaluation script (scripts/evaluation.py) and a dependencies file (scripts/requirements.txt). These are referenced as local files within a project structure, not remote downloads from untrusted sources. Documentation includes standard installation commands for pip packages (anthropic, mcp).
  • [COMMAND_EXECUTION]: Provides example shell commands for running the evaluation script with various transport protocols (stdio, sse, http). These are documented for the user to run manually and are standard for the described purpose.
  • [CREDENTIALS_UNSAFE]: Includes instructions for setting API keys via environment variables (e.g., export ANTHROPIC_API_KEY=your_api_key_here) and placeholders like GITHUB_TOKEN=ghp_xxx. These are standard practices for development and testing, using placeholders rather than hardcoded secrets.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 10:59 AM
Security Audit — agent-trust-hub — model-context-builder