voice-note-ingest
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user-provided audio transcripts and writes them to the local filesystem across several directories.
- Ingestion points: Transcripts are acquired via voice notes or audio attachments processed in SKILL.md.
- Boundary markers: The skill uses block quotes to delimit the "User's Words" section, but the agent's "Analysis" section is instructed to interpret this untrusted content.
- Capability inventory: The skill has permissions to write to multiple directories including originals/, concepts/, people/, companies/, ideas/, personal/, and voice-notes/.
- Sanitization: The skill's "Iron Law" explicitly forbids paraphrasing or cleaning up the transcript, which preserves any malicious instructions embedded by a user within the voice note.
Audit Metadata