Skills
Navigate
skills/omidzamani/dspy-skills/dspy-mcp-tool-integration/Gen Agent Trust Hub

dspy-mcp-tool-integration

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the dspy library with MCP support from a public registry using pip install -U "dspy[mcp]>=3.2.1,<3.3".
  • [COMMAND_EXECUTION]: The documentation includes an example of how to start a local MCP server using the stdio_client interface.
  • Evidence: Uses command="python3" with a placeholder script path in SKILL.md to demonstrate process execution.
  • [PROMPT_INJECTION]: The skill demonstrates ingesting tool definitions from external MCP servers, which serves as an entry point for indirect prompt injection.
  • Ingestion points: The session.list_tools() method call in the example code in SKILL.md.
  • Boundary markers: Not explicitly shown in the simplified integration examples.
  • Capability inventory: Tools retrieved via MCP are converted into dspy.Tool objects and provided to a dspy.ReAct agent for execution.
  • Sanitization: The documentation explicitly recommends applying authentication and authorization as a best practice before exposing tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 10:02 PM
Security Audit — agent-trust-hub — dspy-mcp-tool-integration