follow-up
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data (call notes and transcripts) and incorporate it directly into generated email outputs using exact phrases.
- Ingestion points: External data is ingested via the user-provided call notes and transcripts as described in the prompt examples within
evals/evals.jsonand theDATA FIDELITY RULESinSKILL.md. - Boundary markers: The skill lacks explicit technical delimiters (e.g., XML tags or clear markers) to separate instructions from the data, though it includes a directive to "Extract information ONLY from the call notes provided".
- Capability inventory: The skill is restricted to text generation; it does not utilize tools for file system access, network operations, or subprocess execution.
- Sanitization: There is no evidence of sanitization, escaping, or filtering of the input data before it is interpolated into the email body.
Audit Metadata