ultraqa
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions the agent to dynamically generate and execute temporary scripts, fixtures, and harnesses via shell commands to perform end-to-end behavioral testing.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting user-defined goals and repository content to construct test scenarios.
- Ingestion points: User-provided goal arguments and repository file data.
- Boundary markers: Instructions to use absolute paths and non-interpolating writers for harness generation to prevent unintended shell interpolation.
- Capability inventory: Use of shell tools (e.g., npm, bash) for execution and file writing for temporary artifacts.
- Sanitization: Requirements for safe file writers, isolation of ambient runtime environments (unsetting OMX-style variables), and bounded execution with explicit timeouts.
Audit Metadata