douyin-video

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted audio from external Douyin videos and transcribes it into text. If the video audio contains hidden linguistic instructions designed to manipulate the LLM, the resulting transcript could influence the agent's logic when the output is parsed.
  • [EXTERNAL_DOWNLOADS]: Fetches video files from Douyin's content delivery network (v.douyin.com).
  • [DATA_EXFILTRATION]: Transmits extracted audio files to the SiliconFlow transcription API (api.siliconflow.cn). While this is the intended purpose of the skill, it involves sending local data (audio extracted from downloads) to a third-party service.
  • [COMMAND_EXECUTION]: Invokes the system's ffmpeg binary using the ffmpeg-python library to perform audio extraction, splitting, and metadata probing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 10:10 AM