sticker-forge

AI-assisted die-cut sticker design and production-ready file prep. Takes brand assets as ingredients, composes sticker designs, outputs print-ready PNG + CMYK PDF + SVG cut path for Alibaba manufacturers. Visual QA before delivery.

Model	Source
sonnet	pack: merch

Full Reference

┏━ 🔧 sticker-forge ━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ ┃ [one-line description of sticker design request] ┃ ┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

Overview

AI-assisted design tool that takes brand assets as ingredients, composes sticker designs, and outputs production-ready files for Alibaba sticker manufacturers. NOT a production API integration — a design + file preparation tool. Every output goes through merch-qa visual self-inspection before delivery.

Reference Index

File	Contents
`reference/asset-discovery.md`	Brand asset scanning, folder discovery, asset indexing, color extraction
`reference/file-specs.md`	Resolution, color modes, formats, bleed, safe zones, dimensions
`reference/cut-paths.md`	CutContour standard, vector tracing, path smoothing, offset, SVG export
`reference/materials.md`	Vinyl types, holographic, clear, eco — selection guide per use case

Asset Discovery

Before creating any sticker, discover available brand assets:

Scan current repo for brand assets in common paths: brand/, assets/, public/, static/, images/, src/assets/
Look for SVG, PNG, AI, EPS files
Check for brand.json — if exists, use brand colors and logo paths from it
If no assets found, ask user for folder path
Build asset index: file path, type, dimensions, has-transparency, primary colors

See reference/asset-discovery.md for full scanning algorithm.

Input Modes

Mode	Description	Example
Concept	Describe what you want, sticker-forge composes it from brand assets	”Make a die-cut sticker with our armadillo logo on a holographic-style background with ‘REAL DILLAS’ text”
Reference	Show an example sticker, create something similar with brand assets	”Here’s a sticker I like [image], make something similar with our brand”
Asset	Point to a specific file, prepare as die-cut sticker	”Use this logo [path] as a die-cut sticker, add a white border”
Batch	Multiple designs for a sticker sheet	”Make a sticker sheet with these 6 designs”

Composition Approach

Core principle: Use real brand assets as ingredients — never AI-generate core brand marks.

What AI handles:

Layout and arrangement of brand elements
Background effects, patterns, textures
Border styles and shapes
Text styling and placement
Color harmonies based on brand palette
Supplementary decorative elements around real assets

What AI never does:

Generate the logo from scratch
Modify core brand marks (logos, wordmarks)
Replace brand assets with AI-generated alternatives

When brand.json exists, respect brand colors from it. When extracting colors from assets, use dominant color analysis.

Output Package

Per sticker, output to output/stickers/{name}/:

design.png          # 300 DPI, transparent BG, sRGB
design-cmyk.pdf     # CMYK color profile for offset printing
cutpath.svg         # Vector die-cut line, 1/8" offset from design edge
spec-sheet.md       # Dimensions, bleed, safe zone, material rec
preview.png         # Mockup preview (sticker on surface)

See reference/file-specs.md for detailed format requirements.

Pipeline Flow

1. Asset Discovery
   → Scan repo or use provided path
   → Build asset index
   ↓
2. Design Composition
   → Accept user input (concept/reference/asset/batch mode)
   → Compose design using brand assets as ingredients
   → Generate design.png at 300 DPI, transparent BG
   ↓
3. Production File Prep
   → Generate CMYK PDF for offset printing
   → Trace cut path from design outline
   → Offset cut path 1/8" (3mm) for bleed
   → Export cutpath.svg with CutContour named path
   → Write spec-sheet.md with dimensions and material rec
   ↓
4. Mockup Generation
   → Generate preview.png (sticker on surface)
   ↓
5. Visual QA (merch-qa)
   → Invoke merch-qa with product_type=sticker
   → Pass: deliver output package
   → Fail: revise and re-run QA

QA Gate: Before delivering ANY output, invoke merch-qa with product_type=sticker. Do not present files until QA passes.

Material Recommendations

Material	Best For	Notes
White vinyl matte	Standard indoor stickers	Most common, laptop/water bottle
White vinyl gloss + UV laminate	Outdoor/durable	Weather-resistant, car bumper grade
Holographic film	Premium/collector	Rainbow effect, eye-catching
Compostable vinyl	Eco-conscious brands	Biodegradable, slightly less durable
Clear vinyl	Transparent look	Design must work without white backing

Include material recommendation in spec-sheet.md based on design characteristics.

Integration

Routed by: merch-pimp for all sticker requests
Calls: merch-qa (mandatory before delivery)
Calls: mockup-generator (optional, for preview.png)
Loads: brand.json if present for brand colors/assets