perfect-cuts

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions and scripts involve downloading and installing Python packages (such as whisperx, pyannote-audio, and torch) and Node.js packages (Remotion) from official package registries. These are essential for the skill's primary function of speech recognition and video editing.
  • [COMMAND_EXECUTION]: The skill executes various system commands using the subprocess and spawn modules. It interacts with ffmpeg for video rendering and ffprobe for media metadata extraction. It also uses node to launch the Remotion Studio interface for the user.
  • [PROMPT_INJECTION]: The skill processes transcriptions generated from user-provided video clips. This presents an indirect prompt injection surface where adversarial content within the audio could theoretically attempt to influence the agent's editorial decisions, though risk is mitigated by the agent's structured editorial workflow.
  • [REMOTE_CODE_EXECUTION]: The Remotion launcher script (_remotion-launcher.mjs) performs an npm install to set up a local editing environment and executes npx remotion studio. This behavior is transparently documented and serves the intended purpose of allowing users to preview and refine the automated edits.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 06:46 AM
Security Audit — agent-trust-hub — perfect-cuts