veadk-skills

VeADK Agent Generation

This skill can, according to the user's requirements, generate VeADK Agent code that meets the requirements, or complete VeADK-related functionality.

Trigger Conditions

1. The user briefly describes their functional requirements and wants to build an Agent to accomplish them;
2. The user wants to convert existing Langchain/Langgraph code into VeADK Agent code
3. The user wants to convert an existing Dify workflow into VeADK Agent code

Specific Steps

Below are the different component capabilities of this skill.

Generate an Agent Directly Based on Requirements

Please follow these steps:

1. Analyze the user's requirements and generate the corresponding Agent system structure, refer to references/generator/analyze.md
2. Prompt optimization, refer to references/generator/refine_prompt.md
3. Generate Agent code, refer to references/generator/coding.md

Convert Langchain Code to a VeADK Agent

Please follow these steps:

1. Analyze the original Langchain or Langgraph code
2. Change the original code to a VeADK Agent; for the corresponding relationships, see references/converter/langchain_rules.md
3. Refer to references/common/ the documents in the directory to generate VeADK code

Convert a Dify Workflow to a VeADK Agent

Please follow these steps:

1. Analyze the original Dify workflow DSL (generally a YAML-format file)
2. Change the original code to a VeADK Agent; for the corresponding relationships, see references/converter/dify_rules.md
3. Refer to references/common/ the documents in the directory to generate VeADK code

Follow-up Work

After completing the Agent code writing, call the script to save the code artifacts:

• agent_name/__init__.py: The fixed content is from . import agent # noqa
• agent_name/agent.py: Contains the code for all agents

Among them,agent_name is the name of the Agent you consider appropriate.

The script invocation method is:

python save_file.py --path ... --content ...