The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to read sensitive tokens from ~/.gooseworks/credentials.json and transmit them as URL query parameters (?token=<api_key>) to remote media proxies. This pattern exposes credentials in server logs, proxy logs, and network history.
[EXTERNAL_DOWNLOADS]: The skill fetches external resources from remote locations including Docker images from ghcr.io/gooseworks-ai/goose-video-render and additional skills/scripts via gooseworks fetch <slug> at runtime.
[REMOTE_CODE_EXECUTION]: The instruction set directs the agent to fetch remote 'atoms' or capabilities, run npm install within their directories, and execute the resulting JavaScript files. This creates a vector for remote code execution if the source repository or individual slugs are compromised.
[COMMAND_EXECUTION]: The skill relies on executing powerful local system tools including ffmpeg, ffprobe, and playwright. It also performs shell operations to set environment variables and execute Python scripts that interact with network services.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it processes untrusted data from creative_brief (user-supplied) and extracted_script (template-supplied) without explicit sanitization or boundary markers before using them to drive complex rendering and generation logic.

goose-video