happyhorse-1-0
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the
@runcomfy/clitool to execute video generation workflows. User-supplied prompts are encapsulated within a JSON object passed via the--inputflag, which mitigates traditional shell injection vulnerabilities by avoiding direct shell expansion of user content. - [EXTERNAL_DOWNLOADS]: The skill requires the installation of the
@runcomfy/clipackage from the NPM registry and is hosted in theagentspace-so/runcomfy-skillsGitHub repository. These are standard delivery mechanisms for this vendor's tools. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text prompts that are sent to a remote AI model for video generation. While the skill documented the surface area for indirect injection (specifically regarding image/video inputs in related models), it implements boundaries by treating the input as a string literal within a structured JSON payload. Capability inventory includes executing the vendor CLI, writing generated files to a user-specified directory, and communicating with the
runcomfy.netAPI. Sanitization is primarily handled by the remote model API and the CLI's input handling logic.
Audit Metadata