The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The skill imports a non-standard and unverified dependency 'z-ai-web-dev-sdk'.
Evidence: Found in scripts/chat.ts at line 1.
This package is not from a trusted organization or repository. Third-party SDKs from unknown sources can execute arbitrary code during installation or runtime, posing a supply chain risk.
PROMPT_INJECTION (LOW): The skill exhibits an attack surface for indirect prompt injection by directly interpolating user input into the chat context.
Ingestion points: The prompt parameter in the main function within scripts/chat.ts.
Boundary markers: Absent. User input is passed directly in the message array without delimiters or safety instructions to the model.
Capability inventory: The skill calls a chat completion method (zai.chat.completions.create), which could be manipulated by adversarial input to produce misleading or harmful content.
Sanitization: None. There is no escaping or validation of the prompt string before it is processed by the AI SDK.

LLM