Content Factory

Name/description (YouTube video/Shorts generator) align with what the code and SKILL.md request: FFmpeg and edge-tts for media processing and a Pexels API key for stock footage. Optional features (Whisper, yt-dlp, OpenAI) are documented as optional and appear justified for transcription or advanced LLM features.

SKILL.md instructs the agent to use local binaries, install Python packages, and create a per-skill config directory; the code only reads that config (~/.openclaw-content-factory/config.env) and environment variables limited to an allow-list. Runtime actions (downloading Pexels videos, running edge-tts, ffmpeg operations, optional yt-dlp/whisper) match the documented functionality and do not attempt to read unrelated system files or secrets.

There is no automated install that downloads arbitrary archives; dependencies are Python packages and system binaries documented in the README/SKILL.md. requirements.txt and pip/brew guidance are standard and proportionate to the skill's functionality.

The only required credential is PEXELS_API_KEY (declared as primary). The code optionally respects OPENAI_API_KEY if provided and documents it as optional. The script limits which config keys it loads from the config file, reducing exposure of unrelated secrets.

The skill does not request always: true, does not modify other skills' configs, and only creates a per-user config directory and output folders under the user's home — behavior consistent with a media-generation tool.