tella-auto-cut
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes video transcripts, which are untrusted external data sources, to identify segments for removal. This creates a potential surface for indirect prompt injection where specific text within the video's transcript could attempt to manipulate the agent's logic or trigger unintended editing actions.
- Ingestion points: Transcripts and clip metadata are ingested from the Tella environment via tools like
get_transcriptandget_clipas described inSKILL.md. - Boundary markers: The instructions do not define clear delimiters or specific instructions to the agent to disregard any commands potentially embedded within the transcript text.
- Capability inventory: The skill possesses write capabilities including
cut_clip_by_transcript,cut_clip,remove_fillers,remove_silences, andupdate_clip, which allow it to modify user projects. - Sanitization: No explicit sanitization or validation of the transcript content is performed before the agent uses it to determine cut ranges.
Audit Metadata