The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the official @volcengine/las-cli package from the npm registry. This is the vendor's own command-line interface required for the ASR service.
[COMMAND_EXECUTION]: Executes shell commands via the las-cli tool to manage file uploads to TOS (Torch Object Storage), check service pricing, and initiate transcription tasks.
[CREDENTIALS_UNSAFE]: Includes instructions to run las-cli config show for environment validation. This command typically outputs stored API keys, secret keys, or access tokens, which could lead to accidental exposure of credentials in logs or console output.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted audio and video files provided by the user, creating a surface for indirect prompt injection if the transcribed text contains malicious instructions that the agent might follow.
Ingestion points: External media files processed via the las_asr_pro operator as described in SKILL.md.
Boundary markers: None defined; the transcription result is presented directly to the user or agent.
Capability inventory: Local file system access, network communication with Volcano Engine cloud APIs, and shell command execution via las-cli.
Sanitization: No explicit sanitization or filtering is performed on the transcription text before it is returned to the agent context.

byted-las-asr-pro