image-forge
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The Python scripts within the skill, such as
image_pipeline.pyandbatch_ops.py, perform shell command execution viasubprocess.run. The implementation follows secure coding practices by passing arguments as lists rather than shell-evaluated strings, which mitigates the risk of command injection. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted image files and JSON-formatted specifications. The mandatory evidence chain for this attack surface is as follows: (1) Ingestion points: JSON spec files in
image_pipeline.pyand image paths in all scripts; (2) Boundary markers: Not explicitly used; (3) Capability inventory: Image processing via ImageMagick and background removal tools; (4) Sanitization: Effective use of Python subprocess argument lists. The risk is assessed as safe due to the restricted nature of the tools called. - [EXTERNAL_DOWNLOADS]: The skill uses established dependencies such as
Pillowandrembg. It also leverages thetransparent-backgroundpackage, which is documented to download a research model (InSPyReNet) on its first run, which is a standard procedure for local AI processing.
Audit Metadata