isolated-worker-execution
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is purely instructional and contains no executable scripts, shell commands, or network operations. It is designed to guide an AI agent in helping users build secure sandboxes.
- [NO_CODE]: There are no code files (Python, Node.js, Shell, etc.) provided in the skill package; only Markdown guidelines and YAML evaluation configurations.
- [INDIRECT_PROMPT_INJECTION]: While the skill discusses processing 'less-trusted' data, it explicitly instructs the agent to treat such data as untrusted and to maintain strict boundaries. The evaluation prompts are designed to test the agent's ability to recognize security flaws in hypothetical designs rather than to execute untrusted input.
Audit Metadata