The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads content from several external sources:
scripts/analyze.py uses yt-dlp to download video and audio from user-provided URLs.
scripts/fetch_assets.py fetches images and audio from api.openverse.org, pixabay.com, and freesound.org based on search queries.
scripts/yating_tts.py connects to tts.api.yating.tw for text-to-speech generation.
[COMMAND_EXECUTION]: Extensive use of shell command execution for media processing:
Multiple Python scripts (analyze.py, clip_strip.py, align_lyrics.py, motion_check.py, video.py) invoke ffmpeg and ffprobe for video/audio analysis and manipulation.
scripts/hf_frames.py and scripts/tests/test_engine_render.py execute node and npx commands.
assets/engine-kit/new_project.sh executes shell commands to set up project structures and install Node.js dependencies.
[DYNAMIC_EXECUTION]: The JavaScript runtime component (assets/engine-kit/kit.js) uses eval() to process coordinate expressions provided to the renderSheet function. While intended for mathematical calculations, this allows for dynamic execution of strings.
[CREDENTIALS_UNSAFE]: scripts/yating_tts.py is designed to read a sensitive API key from a fixed path in the user's home directory (~/.reelmimic/secrets.json). While not a hardcoded secret in the skill itself, this pattern identifies where users should store their credentials.

video-clone