draft-in-voice
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an internal knowledge base ('the brain') to generate content, creating a surface for indirect prompt injection.
- Ingestion points: Data is ingested via
gbrain get,gbrain search, andgbrain timelinecalls inSKILL.md, specifically targetingpeople/andmedia/paths. - Boundary markers: The instructions lack explicit delimiters or 'ignore embedded instructions' warnings for the corpus data processed during fingerprint extraction or drafting.
- Capability inventory: The skill has the capability to write to the knowledge base using
gbrain putandgbrain link(as defined in `writes_to: - people/` and the procedure steps).
- Sanitization: There is no evidence of sanitization or filtering of the external content ingested from the brain before it is interpolated into the drafting prompt.
Audit Metadata