The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides a way to process arbitrary user text to generate audio files, which is a known vector for indirect prompt injection.
Ingestion points: User-defined strings in the segments JSON field (SKILL.md).
Boundary markers: Instructions utilize JSON formatting to separate parameters from data.
Capability inventory: Uses the belt tool for remote service interaction.
Sanitization: No specific sanitization logic is provided in the markdown examples.
[EXTERNAL_DOWNLOADS]: The documentation points to external setup guides and CLI tools hosted on the vendor's official GitHub repository. These are legitimate resources required for the skill to function.
[COMMAND_EXECUTION]: The skill demonstrates the use of a CLI tool (belt) to perform tasks like logging in and running dialogue generation applications. These commands are executed by the user to interact with the ElevenLabs API.

elevenlabs-dialogue