verifying-external-behavior
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands including
uv,curl,docker, andunzipto verify the behavior of external systems. While these are intended for development probes, they involve direct interaction with the host system and network. - [INDIRECT_PROMPT_INJECTION]: The skill explicitly discusses verifying behavior from 'scraped documents' and remote API responses. This creates a vulnerability surface where malicious content embedded in those external sources could attempt to influence the agent's behavior.
- Ingestion points: Scraped web documents and remote API response bodies (SKILL.md).
- Boundary markers: None identified in the provided instructions.
- Capability inventory: The skill encourages the use of
uv,curl,docker, and Python to perform system operations (SKILL.md). - Sanitization: No specific sanitization or validation of the external content is mentioned before it is processed or acted upon.
- [DYNAMIC_EXECUTION]: The skill provides a Python snippet that performs runtime monkeypatching of the
requestslibrary (requests.Session.request = spy) to intercept outbound traffic for inspection. This is a form of dynamic code modification for debugging purposes.
Audit Metadata