testing-workflows-with-subagents
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze external data sources, including git logs, error reports, and user feedback, to identify workflow misinterpretations. This represents a vulnerability surface where adversarial content embedded in commit messages or log files could influence the agent's reasoning. The skill does not define specific sanitization or boundary markers for this ingested data.
- [DYNAMIC_EXECUTION]: The workflow requires the agent to generate and execute shell commands (e.g., git repo initialization, branching via git-spice) and dispatch subagents to simulate pressure scenarios. While these are templated operations central to the skill's purpose of setting up test environments, they involve the runtime creation and execution of scripts.
Audit Metadata