The Agent Skills Directory

[COMMAND_EXECUTION]: The helper scripts (e.g., render.py, transcribe.py, grade.py, timeline_view.py) use subprocess.run with list-based arguments to invoke ffmpeg and ffprobe. This approach is secure against shell injection and is necessary for the skill's primary function of video editing.
[EXTERNAL_DOWNLOADS]: The skill communicates with the ElevenLabs API for transcription services and references standard tools like yt-dlp and npx for certain animation workflows. These interactions are with well-known, reputable services and organizations.
[INDIRECT_PROMPT_INJECTION]: The skill has a defined surface for indirect prompt injection as it processes external video files and transcripts.
- Ingestion points: Video data and transcripts enter the context via helpers/transcribe.py.
- Boundary markers: The SKILL.md file mandates strategy confirmation and self-evaluation loops before presenting results to the user.
- Capability inventory: Capability is limited to video manipulation via subprocess calls in helpers/render.py.
- Sanitization: The pipeline is driven by structured JSON Edit Decision Lists (EDL), preventing unsafe string interpolation into shell environments.
[SAFE]: Secret management follows industry standards. The installation guide (install.md) instructs users to store the ElevenLabs API key in a .env file with restricted file permissions (chmod 600), ensuring credentials are not exposed to unauthorized processes.

video-use