Skills
skills/firecrawl/ai-research-skills/dspy/Gen Agent Trust Hub

dspy

Pass

Audited by Gen Agent Trust Hub on Mar 28, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill documents the ProgramOfThought module, which generates and executes Python code. Examples also include a calculate tool using eval() with restricted builtins.\n- [EXTERNAL_DOWNLOADS]: Instructions are provided to install the DSPy library from its official GitHub repository (stanfordnlp/dspy).\n- [PROMPT_INJECTION]: The skill documents systems that ingest data from untrusted sources (vector databases, web search) and possess code-execution capabilities, creating a surface for indirect prompt injection (Category 8). \n
  • Ingestion points: InputField in signatures and Retrieve modules.\n
  • Boundary markers: Not explicitly defined; the framework relies on prompt formatting for separation.\n
  • Capability inventory: Includes code execution via ProgramOfThought and arbitrary tool execution via ReAct.\n
  • Sanitization: Examples show restricted eval() builtins, but general data ingestion lacks validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 28, 2026, 06:07 PM