The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it processes user-provided text through the messages array and transmits it to an external API.
Ingestion points: The messages parameter in the Python code examples within SKILL.md captures user input.
Boundary markers: The skill uses structured role-based dictionaries (e.g., system, user) to separate instructions from user content.
Capability inventory: The skill utilizes the zai library to perform network POST requests to https://open.bigmodel.cn.
Sanitization: No explicit input validation or sanitization is demonstrated for the user content passed into the model.
[CREDENTIALS_UNSAFE]: The documentation identifies a specific local configuration path, ~/.glm/.env, for storing the ZHIPU_API_KEY. While identifying storage locations is part of setup, it highlights a file path containing sensitive credentials.
[EXTERNAL_DOWNLOADS]: The skill references code examples that import the zai Python package, which is not the official Zhipu AI SDK (zhipuai). It also points to documentation and API endpoints hosted at github.com/tencentjerry-afk/zhipu-claude-skills and open.bigmodel.cn.

zhipu-humanoid