wiki-qa
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [SAFE]: The skill defines a standard and safe workflow for codebase exploration. It instructs the agent to search for and read files to answer user queries based on repository evidence.
- [NO_CODE]: This skill contains only instructional markdown and no executable scripts or external dependencies, significantly reducing its attack surface.
- [PROMPT_INJECTION]: The skill identifies an indirect prompt injection surface through the ingestion of repository source files. However, it mitigates risk by instructing the agent to ground answers strictly in evidence and never use external knowledge. * Ingestion points: Codebase files read during analysis in SKILL.md. * Boundary markers: The skill requires inline citations for evidence, which helps distinguish source content from synthesized answers. * Capability inventory: Limited to file searching and reading via repository analysis tools. * Sanitization: Not explicitly defined, but the agent is restricted by grounding rules to prevent following instructions found in data.
Audit Metadata