Claude-skill-registry glif-workflow-builder

Design and build Glif AI workflows. Use when creating new glifs, planning workflow structure, or understanding block types.

install

source · Clone the upstream repo

git clone https://github.com/majiayu000/claude-skill-registry

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/glif-workflow-builder" ~/.claude/skills/majiayu000-claude-skill-registry-glif-workflow-builder && rm -rf "$T"

manifest: skills/data/glif-workflow-builder/SKILL.md

Glif Workflow Builder

Design AI workflows using Glif's block-based system.

Workflow Structure

Glifs are JSON graphs with nodes (blocks) connected via variable references.

{
  "nodes": [
    {"name": "input1", "type": "TextInputBlock", "params": {...}},
    {"name": "generator", "type": "ImageBlock", "params": {"prompt": "{input1}"}}
  ]
}

Block Types

Input Blocks

TextInputBlock - Text input from user
ImageInputBlock - Image upload
MultipickBlock - Multiple choice selector

AI Generation Blocks

TextBlock - LLM text generation (GPT, Claude, Llama)
ImageBlock - Image generation (SDXL, Flux, DALL-E)
ImageToTextBlock - Image captioning/analysis
VideoBlock - Video generation

Utility Blocks

CombineTextBlock - Concatenate text values
JSONExtractorBlock - Parse JSON responses
PublishedGlifBlock - Nest other glifs
WebFetchBlock - HTTP requests

Output Blocks

HTMLBlock - Custom HTML rendering
CanvasBlock - Image composition
AudioBlock - Audio output

Variable References

Reference other blocks using

{blockName}

{
  "name": "enhancer",
  "type": "TextBlock",
  "params": {
    "prompt": "Enhance this prompt: {userInput}",
    "model": "gpt-4"
  }
}

Design Patterns

Basic Image Generator

TextInput → ImageBlock → Output

Prompt Enhancement

TextInput → TextBlock (enhance) → ImageBlock → Output

Multi-Image Composite

TextInput → ImageBlock (bg) ─┐
TextInput → ImageBlock (fg) ─┼→ CanvasBlock → Output

Iterative Refinement

TextInput → ImageBlock → ImageToTextBlock → TextBlock → ImageBlock → Output

Best Practices

Keep inputs minimal - 2-3 max for usability
Name blocks descriptively -
```
stylePrompt
```
not
```
text1
```
Use alphanumeric + underscore only for names
Chain text blocks for complex prompts
Test incrementally - add one block at a time

Schema Reference

Full JSON schema: https://glif.app/api/graphJsonSchema

Common Models

Text: gpt-4, gpt-4o, claude-3, llama-3 Image: flux-pro, sdxl, dall-e-3, stable-diffusion-3 Video: runway-gen3, kling