remove-bg
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on shell execution of a Python script (
skills/shared/scripts/remove_bg.py) to process images with various parameters. - [EXTERNAL_DOWNLOADS]: Upon first run, the skill automatically fetches AI models (such as u2net and isnet) ranging from 4MB to 170MB from remote sources to perform semantic segmentation.
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted image files which may contain embedded instructions in metadata. \n
- Ingestion points: Ingests user-supplied images and external background images via the
-iand--bg-imagearguments inSKILL.md. \n - Boundary markers: No explicit delimiters or instructions are used to tell the agent to ignore potentially malicious content within the processed image files. \n
- Capability inventory: The skill has the capability to execute shell commands and perform filesystem writes within the
outputs/directory. \n - Sanitization: The instruction template uses a placeholder (
outputs/主题名/) for the output path; if the agent interpolates user-controlled strings into this path without proper escaping, it could lead to command injection.
Audit Metadata