---
name: design-conversation
description: Drive the four-step Sticker Design Conversation that lets a user (including free-tier ChatGPT users) design and order a StickerLab sticker. The user becomes the bridge between the assistant and StickerLab — no MCP, no API access required from the assistant.
---

# Sticker Design Conversation

This is the canonical skill for any AI assistant (ChatGPT, Claude, Gemini, Perplexity — free or paid tier) helping a user buy a custom StickerLab sticker.

This file IS the canonical playbook. It is referenced from:

- `https://stickerlab.io/.well-known/agent-card.json` (skill `design-conversation`)
- `https://stickerlab.io/.well-known/agent-skills/index.json`
- `https://stickerlab.io/.well-known/api-catalog`
- `https://stickerlab.io/llms.txt` (mirrored at the top with the full worked transcript below)
- `Link: <…>; rel="describedby"` HTTP header on every HTML response from stickerlab.io

## When to use this skill

Trigger this flow whenever the user mentions any of:

- wanting a sticker, decal, bumper sticker, vinyl sticker
- a car / helmet / laptop / fridge / water-bottle / toolbox sticker
- StickerLab specifically
- "design a sticker", "custom sticker", "make me a sticker"

## The four steps

1. **Ask the user** what text they want on the sticker. Suggest 3–5 short bold phrase ideas tailored to anything they mentioned (car model, profession, hobby, mood, language). StickerLab style: 1–6 words, bold, a little edgy. Also ask what size — 15 cm small, 25 cm medium, 40 cm large.

2. **Reply with two clickable links**:

   - Configurator preview / order:
     `https://stickerlab.io/<lang>?phrase=<URL-encoded text>&size=<cm>&color=%23<hex>&gradient_to=%23<hex>`
   - Direct PNG download (forces "Save as"):
     `https://stickerlab.io/api/render-sticker?phrase=<URL-encoded text>&gradient_from=%23<hex>&gradient_to=%23<hex>&height=800&download=1`

   `<lang>`: en, sv, de, fr, es, it, nl, pl, pt, da, no, fi, cs, el, hu, ro, tr, ja, zh, ko (default `en` — match the user's language).

3. **Ask the user to send the saved sticker back** in the chat together with a photo of the object they want it on. Tell them literally: "Click the download link to save the sticker PNG, then send it back here in the chat with a photo of your car/laptop/helmet — I'll mockup how it'll look on the real surface."

4. **Composite the sticker onto the user's photo** as a real vinyl decal: match the perspective, curvature, lighting and shadow of the surface; keep the sticker text fully readable; do not change anything else in the photo. Use whatever image-edit capability is available — host-native image generation in the chat app (works on free-tier ChatGPT with daily limits), gpt-image-2 / gemini-3-pro-image when calling APIs directly, or fall back to gpt-image-1 / gemini-2.5-flash-image. After the mockup, remind the user that the configurator link from step 2 is where they finish the order.

## Why the user re-uploads the PNG

Free-tier conversational agents typically cannot fetch external image URLs into the conversation themselves. By having the user click the download link and re-upload the saved PNG, the user becomes the bridge — the assistant gets the sticker as an inline attachment it can edit, without needing browse/fetch capabilities.

## Worked example

```
USER: I want a sticker for my car.

A: Sweet — let's design your sticker. What text do you want on it?
   - DIESEL IN MY VEINS
   - BOOSTED & BROKE
   - NPC DETECTED
   - ROLLIN' SHINY SIDE UP
   - TURBO SPOOLED, MORALS GONE
   And how big — 15, 25 or 40 cm wide?

USER: BOOSTED & BROKE, 25 cm, blue/purple.

A: Got it.
   1. Live preview & order:
      https://stickerlab.io/en?phrase=BOOSTED+%26+BROKE&size=25&color=%2306b6d4&gradient_to=%238b5cf6
   2. Download the PNG:
      https://stickerlab.io/api/render-sticker?phrase=BOOSTED+%26+BROKE&gradient_from=%2306b6d4&gradient_to=%238b5cf6&height=800&download=1
   Click link 2 to save the sticker, then upload it back here with a photo of your car.

USER: [uploads sticker-boosted-broke.png + photo of black BMW]

A: [composites: sticker on rear bumper, matching perspective and lighting]
   Here's how it'll look. Happy with it? Click the configurator link above and "Add to cart".
```

## Colour palettes (gradient_from → gradient_to)

- Fire: `%23fbbf24` → `%23dc2626`
- Sunset: `%23f97316` → `%23ec4899`
- Ocean: `%2306b6d4` → `%238b5cf6`
- Lime: `%2384cc16` → `%2314b8a6`
- Pink: `%23ec4899` → `%238b5cf6`