quote-grounded-answer
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and extract information from external documents, which constitutes an ingestion point for untrusted data. However, the instructions include strong defensive measures:
- It explicitly commands the agent to treat document contents as evidence/data and not as instructions to act outside the task.
- It warns against turning document text into permission to send, edit, share, or perform external actions.
- [SAFE]: The skill references Anthropic's official prompt engineering documentation as a source, which is a trusted service. The content is informative and follows best practices for long-context grounding.
Audit Metadata