The Agent Skills Directory

[SAFE]: The skill is a legitimate tool for speech-to-text transcription and alignment, providing instructions on how to use ElevenLabs models via a dedicated platform CLI.
[COMMAND_EXECUTION]: The skill utilizes the belt command-line utility for transcription tasks. The allowed-tools configuration in the skill's frontmatter correctly restricts the agent's execution capabilities to the belt command, adhering to the principle of least privilege.
[EXTERNAL_DOWNLOADS]: The skill references installation guidelines and related platform skills hosted on GitHub under the inference-sh organization. These are standard dependencies required for the skill's intended functionality and originate from well-known service repositories.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to process untrusted external audio and video content.
Ingestion points: Audio and video URLs provided as inputs to the belt CLI commands (SKILL.md).
Boundary markers: None explicitly defined in the instructions; the skill relies on the CLI's handling of the input.
Capability inventory: The skill executes shell commands using the belt CLI to perform inference (SKILL.md).
Sanitization: Inputs are passed as structured JSON objects to the belt command, which is a standard method for preventing simple injection attacks.

elevenlabs-stt