The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a processing surface for untrusted external data from WeChat messages, group notifications, and chat history. \n- Ingestion points: Workflow A (message handling), Workflow D (chat history search), and Workflow C (group management) in SKILL.md.\n- Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when the agent parses message content.\n- Capability inventory: The agent is authorized to forward messages, manage contacts, perform OCR via hotkeys, and append logs to local documentation files (references/self-improving.md).\n- Sanitization: There are no specified sanitization or validation steps for content extracted from the messaging interface.\n- [SAFE]: The skill provides standard documentation for keyboard shortcuts, UI layouts, and troubleshooting steps for the WeChat Windows client. It specifically instructs the agent to defer payment and high-privilege security checks to the user's mobile device, demonstrating a least-privilege approach to desktop operations.

wechat-windows