The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references an external installation guide hosted on GitHub and prompts the installation of the belt-sh/cli skill using the npx skills add command.
[COMMAND_EXECUTION]: The skill uses the belt CLI tool to perform authentication (belt login) and run audio synthesis applications (belt app run) via the system shell.
[DATA_EXFILTRATION]: User-supplied dialogue text is transmitted to the inference.sh service to be processed into audio files.
[INDIRECT_PROMPT_INJECTION]: The skill ingest untrusted dialogue scripts which are then interpolated into shell command arguments, creating a surface for potential indirect prompt injection attacks.
Ingestion points: The segments input object in SKILL.md which contains user-defined text and voice parameters.
Boundary markers: The input is encapsulated in a JSON structure, but the skill lacks explicit instructions for the AI to ignore or sanitize instructions embedded within the dialogue text segments.
Capability inventory: Shell command execution via Bash(belt *) and the ability to write output to local files (e.g., dialogue.json).
Sanitization: No sanitization or escaping mechanisms are defined for the user-provided text segments before they are passed to the CLI.

elevenlabs-dialogue