edit-video
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes system binaries including
ffmpeg,ffprobe, and a localwhisper-clibuild using theBun.spawnAPI. These operations are restricted to video processing tasks such as audio extraction, segment cutting, and concatenation, which are the primary functions of the skill. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it transcribes audio from external video files and presents the text to the agent for analysis and planning.
- Ingestion points: External video files are processed via
scripts/transcribe.tsandscripts/lib/whisper.ts. - Boundary markers: The transcript is formatted into markdown tables in
scripts/lib/transcript.tsandscripts/lib/analyze.tswithout specific boundary markers or instructions to ignore embedded commands. - Capability inventory: The agent can read and write files on the local filesystem and execute video processing commands via
ffmpeginscripts/lib/ffmpeg.ts. - Sanitization: Transcribed text is not sanitized or filtered for potential malicious instructions before being included in the agent's context.
Audit Metadata