The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data via a 'prompt' or 'brief' input which is interpolated into API requests sent to ElevenLabs and Fal AI. There are no explicit boundary markers or instructions to the model to ignore embedded commands within this user-supplied content, creating a surface for indirect prompt injection.
[COMMAND_EXECUTION]: The skill's primary functionality is implemented through shell scripts (e.g., scripts/compose.sh, scripts/video_to_music.sh) that use curl to perform network operations and jq for JSON processing. These scripts are used to send data and credentials to external AI services.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of generated audio files from the official ElevenLabs API (api.elevenlabs.io) and Fal AI (fal.run). These are well-known technology providers, and the downloads are consistent with the skill's stated purpose.
[DATA_EXPOSURE]: The SKILL.md file contains documentation demonstrating how to programmatically extract credentials from a .env file using shell commands. While presented as a fallback instruction, it highlights the skill's reliance on environment-stored secrets.
[DYNAMIC_EXECUTION]: The script scripts/compose_detailed.sh utilizes python3 - to execute a block of Python code passed via standard input. This code is used for parsing multipart MIME responses from the ElevenLabs API.

create-music-elevenlabs