The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_audio.py uses the subprocess.run function to interact with ffmpeg for audio processing tasks, including determining audio duration, generating silent segments, and concatenating multiple audio files.
Evidence: Subprocess calls are used for FFmpeg operations with arguments passed as a list, which prevents shell injection. The binary path is configurable via the FFMPEG_PATH environment variable or defaults to a project-local bin directory.
[EXTERNAL_DOWNLOADS]: The skill connects to the ElevenLabs API (api.elevenlabs.io) to transform narration text into audio files. This is a well-known third-party service for Text-to-Speech.
[PROMPT_INJECTION]: The skill processes untrusted text data from a scenes.json file (specifically the narration field). While this text is passed to a TTS engine and not an LLM, it represents a surface for indirect prompt injection if the source data is attacker-controlled.
Ingestion points: scenes.json narration field.
Capability inventory: Local file writing (.mp3, .json), network communication via urllib.request, and local command execution via FFmpeg.
Sanitization: The skill performs no specific sanitization on the input text before sending it to the TTS API, which is standard for this utility's purpose.

video-audio