The Agent Skills Directory

[SAFE]: The skill is a standard API connector authored by a trusted vendor (Zapier). All network operations are directed to ElevenLabs' official API endpoints (api.elevenlabs.io) and Zapier's infrastructure.
[COMMAND_EXECUTION]: The cli.js file uses node:child_process.spawnSync to execute the connector's own logic (either cli.ts or the compiled dist/cli.js). This is a common pattern for providing CLI access to Node.js applications and does not involve executing arbitrary user-provided commands.
[EXTERNAL_DOWNLOADS]: Tools like speechToText and isolateAudio allow the agent to fetch audio data from user-provided HTTPS URLs for processing. This is a core feature of the skill for handling remote audio assets.
[DATA_EXPOSURE]: The skill includes functionality to read from local file paths (via the audio_path input) and write generated audio to the system's temporary directory. These operations are intended for managing audio artifacts during the generation and transformation process.
[PROMPT_INJECTION]: The SKILL.md includes safety guidelines for the agent, such as requiring user clarification when multiple voices match a name search, preventing unintended actions on the wrong resources.
[CREDENTIALS_UNSAFE]: The skill uses environment variables (env:) for direct API key storage, which is a recommended practice. The documentation specifically advises the agent not to leak these credentials by using safe command variants (e.g., env | cut).

elevenlabs