offensive-cloud
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to ingest and analyze untrusted data, including cloud configuration files, source code, and network traces. 1. Ingestion points: SKILL.md specifies reading configuration files, source search, and request/response traces. 2. Boundary markers: There are no instructions to use delimiters or ignore embedded instructions within the ingested data. 3. Capability inventory: The workflow includes executing npm test, toolchain builds, and writing/executing verification commands. 4. Sanitization: No sanitization or validation of the analyzed content is mentioned.
- [DYNAMIC_EXECUTION]: The workflow instructs the agent to generate and execute remediation code and automated regression tests (e.g., npm test) based on its analysis of external data. This runtime generation and execution of code is a sensitive capability.
Audit Metadata