agentsop-code-execution-decision
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructs agents to dynamically generate and execute Python code to solve deterministic problems like arithmetic and data manipulation. It explicitly addresses the security risks of this behavior by recommending the use of sandboxed environments (containers or VMs) and warning against the use of unsandboxed tools like LangChain's PythonREPLTool for untrusted inputs.
- [INDIRECT_PROMPT_INJECTION]: The skill outlines workflows for processing user-supplied data such as CSVs or financial line items. It recognizes the potential for malicious data to influence agent behavior and mitigates this risk by advocating for strict output formatting, execution sandboxing, and human-in-the-loop narration of code results.
- [EXTERNAL_DOWNLOADS]: The skill provides numerous references to well-known academic research papers on ArXiv and official documentation for industry-standard AI tools (OpenAI, Anthropic, LangChain, DSPy). These external links are used solely for provenance and educational purposes and do not represent a remote code execution risk.
Audit Metadata