image-inpainting
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
runcomfyCLI via Bash to perform image processing tasks. The scope is restricted to the specific CLI tool, and user-provided inputs are passed via a JSON string to prevent shell command injection. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied image and mask URLs, which represents a potential surface for indirect prompt injection via image content or metadata. The skill documentation explicitly acknowledges this risk and advises users to only process trusted URLs.
- Ingestion points: User-provided image and mask URLs in
SKILL.md. - Boundary markers: Inputs are delimited within a JSON structure passed to the CLI.
- Capability inventory: Execution of the
runcomfyCLI tool which makes network requests to the vendor's API. - Sanitization: The CLI avoids shell expansion by parsing input as structured data.
Audit Metadata