zhipu-image-generation
Warn
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill identifies a specific sensitive file path (
~/.glm/.env) for storing and loading theZHIPU_API_KEY. Instructing an agent to access hidden configuration files increases the risk of accidental exposure of credentials. - [EXTERNAL_DOWNLOADS]: The SDK usage example references a Python package named
zai. The official library for Zhipu AI iszhipuai. The use of a non-standard or potentially typosquatted package name poses a security risk, as it may lead to the installation of unverified or malicious code from public registries. - [INDIRECT_PROMPT_INJECTION]: The skill accepts user-supplied data in the
promptfield for image generation. This content is passed directly to an external API (open.bigmodel.cn), creating an attack surface where malicious instructions embedded in the user prompt could influence the agent's behavior or the API's execution context.
Audit Metadata