aikey-cut
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes untrusted transcript data from video footage.
- Ingestion points: Video transcripts are read via read_script, read_captions, or local transcription tools.
- Boundary markers: The skill lacks technical delimiters for transcript content but provides instructional boundaries (e.g., forbidding the addition of words not in the original speech).
- Capability inventory: The skill utilizes toolsets including apply_script for timeline modification, submit_export for file rendering, and chatcut-video-gen for generative AI content.
- Sanitization: The skill mandates the masking of sensitive data such as revenue and personal identity in video frames and subtitles using solid color blocks.
Audit Metadata