glm-image-gen
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script
scripts/glm_image_cli.pyto communicate with the ZhiPu GLM-Image API. This is the primary intended mechanism of the skill.\n- [EXTERNAL_DOWNLOADS]: The skill downloads images from ZhiPu AI's official domain (open.bigmodel.cn) when the--saveparameter is used. As this targets a well-known service infrastructure, it is documented as a standard functional requirement.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for indirect prompt injection as it processes user-provided text to generate images. \n -
- Ingestion points: The
--promptargument inscripts/glm_image_cli.py.\n
- Ingestion points: The
-
- Boundary markers: The skill instructs the agent to pass user-provided descriptions directly to the CLI script.\n
-
- Capability inventory: The script performs network POST requests to the API, network GET requests to download images, and local file writes when the
--saveflag is specified.\n
- Capability inventory: The script performs network POST requests to the API, network GET requests to download images, and local file writes when the
-
- Sanitization: The script relies on the ZhiPu API's built-in
content_filterto detect and block inappropriate prompt content, and the script includes logic to parse and report these safety levels.
- Sanitization: The script relies on the ZhiPu API's built-in
Audit Metadata