video-subtitle-cutter

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes subprocess.run to execute ffmpeg, ffprobe, and whisper commands locally. While it uses list-based arguments to mitigate shell injection, these operations interact directly with the local file system and system resources.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the openai-whisper Python package and the ffmpeg system utility. It also makes network requests to the OpenAI Whisper API (api.openai.com) for transcription services. These are well-known services and expected for the skill's core functionality.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8). It ingests untrusted transcription data generated from user-provided videos and passes it to an LLM to identify video segments. A malicious transcript could attempt to override the LLM's instructions to influence which parts of the video are kept or removed.
  • Ingestion points: Reads transcript.json generated from video audio.
  • Boundary markers: Uses TRANSCRIPT: and TRANSCRIPT SEGMENTS: headers but lacks strict delimiters or instructions to ignore embedded commands.
  • Capability inventory: Includes file read/write (open, write), file deletion (os.remove), and subprocess execution (ffmpeg).
  • Sanitization: No explicit sanitization or validation is performed on the text content of the transcript before LLM processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 02:58 AM