The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions specify installing the belt CLI tool using npx skills add belt-sh/cli and provide a link to installation documentation hosted on GitHub. These references point to the official repository for the CLI tool used by the skill.
[COMMAND_EXECUTION]: The skill utilizes a set of belt commands to manage authentication and execute audio processing applications. These commands are run via a Bash tool, with access restricted to the belt binary through the allowed-tools configuration.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external URLs (audio/video files) to generate transcripts. These transcripts are returned to the agent's context, creating a potential surface where malicious instructions embedded in the audio content could influence the agent's subsequent actions.
Ingestion points: The audio and video_url input fields specified in the belt app run examples.
Boundary markers: Command inputs are structured as JSON, but the skill does not define specific delimiters or "ignore instructions" warnings for the agent when handling the resulting transcription output.
Capability inventory: The agent can execute shell commands via the belt CLI, including logging into the service and running inference applications.
Sanitization: No explicit sanitization, validation, or filtering of the processed audio/video content or the resulting text transcripts is described in the skill definition.

elevenlabs-stt