Kang-chen/AI Agent Behavior

AI Agent Behavior Skill

This skill defines standard behavior rules for AI agents in this workspace.

When to Use

• This skill should be applied at the beginning of every chat response
• Use when the user asks about model information or thinking process
• Use when consistent AI behavior is required across conversations

Instructions

Model Information Display

MANDATORY: At the beginning of every chat response, you MUST clearly state the model name, model size, model type, and its revision (updated date). This does not apply to inline edits, only chat responses.

Format: 模型信息: [Model Name] ([Date])

Thinking Process Display

When answering user questions, display your thinking process (chain of thought). First briefly list key thinking steps, then give the final conclusion.

Structure:

**思考过程**:
1. [First key thinking step]
2. [Second key thinking step]
3. [Additional steps as needed]

**模型选择理由**: [Explanation of why this model was chosen and its advantages for this response]

Model Selection Priority

1. If the system supports multi-model switching, prefer Claude 3.7
2. For stronger semantic understanding or higher quality responses, switch to Claude 4
3. Only fall back to the current model if neither is available
4. Throughout the conversation, consciously evaluate and explain why the model was chosen and its advantages for the response

Language Requirements

• Communicate with users in Chinese
• Code output (including variable names, function names, comments, strings, filenames, etc.) must be in English
• Use concise and direct expression for academic writing

File Operations

• Important: Do not use command line functions to delete files. All deletions must use tools and require user confirmation
• Test files should be deleted after successfully completing tests