The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data through video files, URLs, and subtitle scripts.
Ingestion points: video_path, video_url, and subtitles parameters.
Boundary markers: The instructions lack explicit prompt delimiters to isolate content from these external sources when the agent processes them.
Capability inventory: The skill has access to the Bash tool and file system Write operations, which could be influenced if malicious instructions in subtitles are interpreted by the agent.
Sanitization: The backend performs validation of subtitle formats and language consistency, which serves as a mitigation against simple malformed inputs.
[COMMAND_EXECUTION]: The skill invokes the sonilo CLI via the Bash tool to perform dubbing tasks and poll for status. This is a core feature of the skill used to interact with the vendor's service.
[EXTERNAL_DOWNLOADS]: The skill fetches video content and subtitle files from user-provided HTTPS URLs. These operations target vendor-specified or user-provided endpoints required for the dubbing service.

auto-dubbing