The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill converts untrusted audio and video content into text transcripts which are then returned to the agent's context. If the source media contains crafted speech with malicious instructions, it could lead to indirect prompt injection.
Ingestion points: Audio and video files provided via local file uploads or remote URLs in the cloud_storage_url parameter.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are included in the transcription instructions.
Capability inventory: The skill produces plaintext transcripts and speaker identification data for agent processing.
Sanitization: No sanitization or validation of the transcribed text is implemented within the skill logic.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of several official libraries from public registries (NPM and PyPI).
Python packages: elevenlabs, python-dotenv, pydub.
Node.js packages: @elevenlabs/elevenlabs-js, @elevenlabs/client, @elevenlabs/react, dotenv.
Observation: The skill documentation proactively guides users away from a deprecated legacy package in favor of the official scoped @elevenlabs namespace.

speech-to-text