The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data via the voice_url and music_url parameters, which creates a potential surface for indirect prompt injection. If the processed media contains spoken instructions that are subsequently transcribed by the agent, they could influence the agent's behavior.
Ingestion points: The audio_ducking tool accepts remote HTTPS URLs through the voice_url and music_url parameters.
Boundary markers: There are no explicit instructions or delimiters used to ensure the agent ignores any commands embedded within the audio or video content.
Capability inventory: The skill possesses the ability to make network requests to the Sonilo API (api.sonilo.com) and write resulting files to the local filesystem (output_directory).
Sanitization: No sanitization or validation mechanisms are mentioned to filter or check the content of the external media files before they are processed.

audio-ducking