image-to-video
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the
runcomfy runcommand via the local system shell to perform image-to-video animation tasks. It passes user-supplied parameters like prompts and URLs as a JSON string to the CLI tool. - [EXTERNAL_DOWNLOADS]: The skill requires the installation of the
@runcomfy/clipackage from the npm registry to function. This is a legitimate dependency for interacting with the provider's infrastructure. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources via
image_url,video_url, andaudio_urlparameters. - Ingestion points: Media URLs provided in the JSON input in
SKILL.md. - Boundary markers: The skill uses JSON formatting to isolate inputs for the CLI but does not specify delimiters for the AI model processing the media.
- Capability inventory: Execution of the
runcomfyCLI and file downloads to a user-specified--output-dir. - Sanitization: The skill warns that image-based prompt injection is a risk for the model server but does not provide specific local sanitization for the media contents.
- [CREDENTIALS_UNSAFE]: Authentication is managed through
runcomfy login, which stores the API token in~/.config/runcomfy/token.json. The documentation states that the file is created with restricted mode 0600 permissions, following standard security practices for CLI credential storage.
Audit Metadata