The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides code and instructions for executing system commands. Examples include the use of ffmpeg and ffprobe for video processing in rules/ffmpeg.md and rules/transcribe-captions.md. Additionally, the transcription rule involves running compiled whisper.cpp binaries.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download and installation of external software and assets. In rules/transcribe-captions.md, it provides logic to install whisper.cpp and associated AI models. rules/sfx.md directs users to a third-party GitHub repository (github.com/kapishdima/soundcn) for sound effects. Several rules also involve fetching data from external URLs (e.g., Lottie animations and Mapbox styles).
[CREDENTIALS_UNSAFE]: The skill requires sensitive API keys for external services. rules/voiceover.md necessitates an ELEVENLABS_API_KEY and instructs the agent to ask the user for it if missing. rules/maps.md requires a REMOTION_MAPBOX_TOKEN to be stored in a .env file.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by processing untrusted data.
Ingestion points: Data is fetched from user-supplied URLs in rules/calculate-metadata.md and JSON caption files are loaded in rules/display-captions.md. Audio files are also processed for transcription in rules/transcribe-captions.md.
Boundary markers: The code snippets do not implement clear delimiters or safety warnings to prevent the agent from following instructions embedded in processed data.
Capability inventory: The skill uses network requests (fetch), file system writes (writeFileSync), and shell command execution (execSync).
Sanitization: There is no evidence of validation or sanitization for the content retrieved from external URLs or parsed from input files.

remotion-best-practices