rjsamra/dify-skills

Dify Skills

Agent skills for generating Dify workflow DSL YAML that can be imported directly into Dify.

Repository: github.com/rjsamra/dify-skills

Install

Works with Cursor, Claude Code, Codex, OpenCode, and 40+ other agents via the skills CLI:

npx skills add rjsamra/dify-skills

Useful variants:

# List available skills
npx skills add rjsamra/dify-skills --list

# Install only this skill, globally, non-interactive
npx skills add rjsamra/dify-skills --skill dify-creator -g -y

# Install to specific agents
npx skills add rjsamra/dify-skills -a cursor -a claude-code

Claude Code plugin (optional)

/plugin marketplace add rjsamra/dify-skills
/plugin install dify-creator@dify-skills

Available Skills

dify-creator

Guide users through multi-turn conversation to clarify requirements, reference 125+ existing Dify cases, and generate importable workflow DSL YAML.

Use when:

• Building a Dify workflow, chatflow, or advanced chat app
• Generating Dify DSL YAML from a product brief
• Matching a use case to known Dify patterns (RAG, OCR, TTS, agents, etc.)

Directory Structure

dify-skills/
├── README.md
├── .claude-plugin/
│   └── marketplace.json       # Claude Code plugin marketplace
└── skills/
    └── dify-creator/
        ├── SKILL.md           # Skill instructions
        └── organized_dsl/     # Reference Dify DSL case library
            ├── INDEX.md
            ├── Dify_DSL_Complete_Node_Reference_Guide.md
            ├── 01_content_generation_and_creation/
            ├── 02_image_generation_and_design/
            ├── ...
            └── 14_reference_examples/

Overview

dify-creator helps agents quickly create Dify-compatible workflow configs. Through guided conversation and reference to existing cases, it produces YAML files that can be imported directly into the Dify platform.

Features

• Guided conversation: Multi-turn Q&A to clarify requirements
• Learning from examples: Auto-match best practices from existing Dify cases
• Full DSL generation: Output complete configs that follow Dify conventions
• Multi-type support: Chatflow, Workflow, and Advanced Chat
• Variable tracking: Manage variable passing and references between nodes

Quick Start

After installing the skill, ask your agent:

I want to create a Dify workflow for meeting minutes from audio

The skill will guide you through:

1. Confirm basic info (name, description, app type)
2. Understand functional needs (input, processing, output)
3. Analyze flow structure (branches, loops, error handling)
4. Choose models and tools
5. Reference existing cases under organized_dsl/
6. Generate a complete DSL config

Application Types

Type	Description	Typical use
workflow	Batch tasks, single-run	Data processing, batch generation
chatflow	Conversational app, multi-turn	Customer service, chat assistants
advanced-chat	Advanced chat mode	Complex conversational scenarios

Supported Node Types

Node	Function
start	Start node; receive user input
llm	Large language model call
tool	Tool/plugin call
answer	Reply to the user
if-else	Conditional branching
code	Code execution
iteration	Loop processing
template-transform	Template transform
variable-aggregator	Variable aggregation
knowledge-retrieval	Knowledge base retrieval

Best Practices

1. Clarify requirements: Describe the workflow goal clearly in conversation
2. Reuse patterns: Ask the agent to reference similar existing cases
3. Iterate: Build the core flow first, then add branches and error handling
4. Validate: Import into Dify and debug/optimize

FAQ

Q: Are custom plugins supported?
A: Yes. Describe the plugins you need; the skill will declare them under dependencies.

Q: Can it generate Chatflow?
A: Yes. Specify the app type as chatflow or advanced-chat in conversation.

Q: What is the variable reference format?
A: Use {{#nodeID.outputField#}}.

License

This project is licensed under CC BY-NC-SA 4.0.

Author

Rajkumar — rjsamra/dify-skills