Community编程与开发github.com

SystemTce/dify-skills

Dify 工作流和插件开发的完整技能体系 - 模块化、渐进式、实用导向

dify-skills 是什么?

dify-skills is a Claude Code agent skill that dify 工作流和插件开发的完整技能体系 - 模块化、渐进式、实用导向.

兼容平台✓Claude Code✓Codex CLI~Cursor
npx skills add SystemTce/dify-skills

Installed? Explore more 编程与开发 skills: steipete/bluebubbles, steipete/eightctl, steipete/blucli · View all 6 →

在你喜欢的 AI 中提问

打开一个已预加载此 Agent Skill 的新对话。

文档

Dify Master SKILL

A modular, progressively disclosed Dify development skill system

Quick Navigation


What is Dify

Dify is an open-source LLMOps platform that provides visual workflow orchestration and a powerful plugin system, helping developers quickly build and deploy AI applications.

Core Features

  • Visual Workflows: Drag-and-drop orchestration, build complex AI apps without coding
  • Plugin System: Supports multiple plugin types: tools, models, data sources
  • Production Ready: Complete support from local development to large-scale distributed deployment
  • Open Source Ecosystem: 114K+ GitHub Stars, active community, and rich plugin marketplace

Technical Architecture

  • Beehive Architecture: Modular, extensible system design
  • Queue-Driven Graph Engine: Event-driven, parallel execution, high performance
  • Multi-Runtime Support: Local, Debug, and Serverless operating modes
  • MCP Protocol Integration: Standardized integration with AI assistants like Claude

SKILL System Overview

Dify Master SKILL adopts a modular design with 6 core sub-skills covering the complete development-to-deployment pipeline.

dify-master (main entry)
├── 01-workflow      # Workflow design
├── 02-plugin        # Plugin development
├── 03-performance   # Performance optimization
├── 04-security      # Security and deployment
├── 05-integration   # Integration and extension
└── 06-reference     # Reference materials

Design Philosophy

  1. Modular: Each sub-skill is independent and complete, usable on its own
  2. Progressive Disclosure: From overview to details, go deeper as needed
  3. Context Optimized: Minimize the amount of content loaded per session
  4. Practical Oriented: Provide ready-to-use code and configurations

Quick Start

Scenario 1: Create Your First Workflow

# Simple Q&A workflow
workflow:
  graph:
    nodes:
    - data:
        type: start
        variables: []
      id: start

    - data:
        type: llm
        model:
          provider: openai
          name: gpt-4
        prompt_template:
        - role: system
          text: "You are an assistant"
        - role: user
          text: "{{#sys.query#}}"
      id: llm

    - data:
        type: answer
        answer: "{{#llm.text#}}"
      id: answer

    edges:
    - source: start
      target: llm
    - source: llm
      target: answer

Next step: Learn 01-workflow SKILL for more node types and design patterns

Scenario 2: Develop Your First Plugin

# Simple tool plugin
from typing import Any, Generator
from dify_plugin.entities.tool import ToolInvokeMessage
from dify_plugin import Tool

class MyTool(Tool):
    def _invoke(
        self,
        tool_parameters: dict[str, Any]
    ) -> Generator[ToolInvokeMessage, None, None]:
        # Get parameters
        input_text = tool_parameters.get("text", "")

        # Processing logic
        result = input_text.upper()

        # Return result
        yield self.create_text_message(result)

Next step: Learn 02-plugin SKILL for the complete plugin development workflow

Scenario 3: Optimize Workflow Performance

# Use parallel branches to boost performance
- data:
    type: parallel
    branches:
    - nodes: [tool1, process1]
    - nodes: [tool2, process2]
    - nodes: [tool3, process3]
  id: parallel_execution

Next step: Learn 03-performance SKILL for more optimization techniques


Sub-SKILL Index

01. Workflow Design SKILL

Use Cases: Design and develop Dify workflows

Trigger Keywords: node, LLM, Agent, workflow, variable, knowledge retrieval, HTTP Request, Code, If-Else, Iteration, template, DSL, YAML

Core Content:

  • 10+ node types explained (Start, LLM, Agent, Knowledge Retrieval, Code, HTTP Request, Tool, If-Else, Iteration, Answer)
  • 5 design patterns (Sequential Pipeline, Conditional Branching, Iterative Processing, Knowledge Retrieval, Multi-Source Aggregation)
  • Ready-to-use workflow templates
  • Variable system and reference syntax
  • Best practices and common issues

Quick Access: 01-workflow/SKILL.md

Recommended for: Workflow designers, AI application developers, product managers


02. Plugin Development SKILL

Use Cases: Develop custom tool, model, and data source plugins

Trigger Keywords: plugin, Tool, Provider, development, manifest, OAuth, data source, model, extension

Core Content:

  • 4 plugin types (Tool, Model, Datasource, Extension)
  • Complete development workflow (initialize, implement, test, package, publish)
  • Standard project structure and configuration files
  • Provider and Tool development guides
  • Plugin templates and code examples
  • Testing strategies and debugging tips

Quick Access: 02-plugin/SKILL.md

Recommended for: Plugin developers, backend engineers, tool integrators


03. Performance Optimization SKILL

Use Cases: Optimize execution performance of workflows and plugins

Trigger Keywords: performance, optimization, cache, parallel, latency, throughput, monitoring, benchmark

Core Content:

  • Workflow optimization (graph structure, parallel execution, variable pool)
  • LLM optimization (model selection, prompt engineering, caching, structured output)
  • Plugin optimization (runtime selection, resource limits, async processing)
  • Monitoring and analysis tools
  • Performance benchmarks

Quick Access: 03-performance/SKILL.md

Recommended for: Performance engineers, architects, operations staff


04. Security and Deployment SKILL

Use Cases: Production environment deployment and security hardening

Trigger Keywords: security, deployment, Docker, Kubernetes, SSL, authentication, authorization, sandbox, SSRF

Core Content:

  • Security practices (SSRF protection, sandbox isolation, data encryption, authentication/authorization)
  • Deployment guides (Docker Compose, Kubernetes, Serverless, high availability)
  • Operations practices (monitoring, logging, backup, troubleshooting)
  • Environment configuration and resource planning
  • Best practices and case studies

Quick Access: 04-security/SKILL.md

Recommended for: DevOps engineers, security engineers, system administrators


05. Integration and Extension SKILL

Use Cases: Integrate with external systems and services

Trigger Keywords: integration, API, Webhook, MCP, Server, Client, database, Redis, message queue, RabbitMQ, Kafka, S3

Core Content:

  • API integration (REST API, Webhook, Streaming)
  • MCP protocol (MCP Server, MCP Client, bidirectional communication)
  • External services (databases, storage, message queues)
  • Integration case studies and best practices
  • Troubleshooting and debugging

Quick Access: 05-integration/SKILL.md

Recommended for: Integration engineers, full-stack developers, architects


06. Reference SKILL

Use Cases: Look up syntax, API, CLI, and other reference information

Trigger Keywords: reference, API, SDK, CLI, DSL, YAML, command, documentation, syntax

Core Content:

  • DSL syntax reference (complete YAML structure and field descriptions)
  • API reference (RESTful API interface documentation)
  • SDK reference (Python SDK classes and methods)
  • CLI reference (dify command line tool)
  • Common questions and solutions
  • Learning resources and community links

Quick Access: 06-reference/SKILL.md

Recommended for: All Dify developers


Common References

Variable Reference Syntax

# System variables
{{#sys.query#}}              # User query
{{#sys.files#}}              # Uploaded files
{{#sys.conversation_id#}}    # Conversation ID

# Node output
{{#nodeID.outputField#}}     # Reference node output
{{#llm.text#}}               # LLM text output
{{#knowledge.result#}}       # Knowledge retrieval result

# Conversation variables
{{#conversation.variableName#}}  # Conversation-level variable

# Iteration variables
{{#item#}}                   # Current iteration item
{{#index#}}                  # Current index

Common Node Configurations

# LLM node
- data:
    type: llm
    model:
      provider: openai
      name: gpt-4
      completion_params:
        temperature: 0.7
        max_tokens: 2000
    prompt_template:
    - role: system
      text: "System prompt"
    - role: user
      text: "{{#sys.query#}}"
    context:
      enabled: true
      variable_selector: ["knowledge", "result"]

# Agent node
- data:
    type: agent
    agent_parameters:
      instruction:
        value: "Call tools based on user requirements"
      model:
        value:
          provider: openai
          model: gpt-4
      query:
        value: "{{#sys.query#}}"
      tools:
        value:
        - enabled: true
          provider_name: time
          tool_name: current_time
          type: builtin
    agent_strategy_name: function_calling

# Knowledge retrieval node
- data:
    type: knowledge-retrieval
    dataset_ids: ["dataset-id"]
    query_variable_selector: ["start", "sys.query"]
    retrieval_mode: single
    single_retrieval_config:
      model:
        provider: openai
        name: text-embedding-3-small
      top_k: 3
      score_threshold: 0.7

Plugin Configuration Template

# manifest.yaml
version: 0.0.1
type: plugin
author: your-name
name: plugin-name
label:
  en_US: Plugin Name
  zh_Hans: Plugin Name
description:
  en_US: Plugin description
  zh_Hans: Plugin description
icon: icon.svg

resource:
  memory: 268435456  # 256MB

plugins:
  tools:
    - provider/provider.yaml

meta:
  version: 0.0.1
  arch: [amd64, arm64]
  runner:
    language: python
    version: "3.12"
    entrypoint: main

tags:
  - utilities

Common CLI Commands

# Plugin development
dify plugin init                    # Initialize a plugin project
dify plugin run                     # Debug plugin locally
dify plugin package                 # Package plugin
dify plugin publish                 # Publish plugin

# Workflow management
dify workflow export <id>           # Export a workflow
dify workflow import <file>         # Import a workflow
dify workflow validate <file>       # Validate a workflow

# Environment management
dify env list                       # List environment variables
dify env set <key> <value>          # Set environment variable
dify env get <key>                  # Get environment variable

Learning Path

Beginner Path (0-1 Week)

  1. Read this SKILL to understand the Dify landscape
  2. Learn 01-workflow SKILL to master workflow fundamentals
  3. Create your first workflow using a template
  4. Test and debug on the Dify platform

Intermediate Path (1-4 Weeks)

  1. Deep-dive into 01-workflow SKILL design patterns
  2. Learn 02-plugin SKILL to develop custom plugins
  3. Learn 03-performance SKILL to optimize performance
  4. Build and deploy a real project

Expert Path (1-3 Months)

  1. Master all sub-skill content
  2. Learn 04-security SKILL for production deployment
  3. Learn 05-integration SKILL to integrate external systems
  4. Contribute to the community and share best practices

Getting Help

Official Resources

Community Resources

SKILL Support

  • Architecture Document: ARCHITECTURE.md
  • Issue Feedback: Submit via GitHub Issues
  • Contribution Guide: Help improve and extend SKILL content

Version Information

  • Current Version: v1.1.0
  • Last Updated: 2026-03-05
  • Compatible Dify Versions: 0.15.x - 1.9.x
  • Update Content: Completed 05-integration and 06-reference SKILLs, added on-demand loading mechanism
  • Next Version Plan: v1.2.0 (add more templates and case studies)

License

This SKILL is open-sourced under the MIT License. You are free to use, modify, and distribute it.


Start your Dify development journey! Choose a sub-SKILL to dive deeper, or jump directly into templates to get started quickly.

Individual skills in this repo

This repo contains 6 individual skills — each has its own dedicated page.

相关技能