Skills

gpt-image-2-gen

Installation
SKILL.md

GPT Image 2 Generation

An interactive AI image generation assistant powered by the GPT Image 2 model via EvoLink API.

When to Activate This Skill

Activate this skill when the user asks to:

  • Generate / create / make an image or picture
  • Edit / modify an existing image
  • Use GPT Image 2 or gpt-image-2
  • Create AI art, illustrations, logos, icons, or any visual content
  • Batch-generate image variations

Keywords: image, picture, illustration, photo, art, logo, icon, generate image, create image, image editing, gpt-image, text-to-image

Script Location

IMPORTANT: All script paths in this file are relative to the directory containing this SKILL.md file. Before running any script, resolve the absolute path:

SKILL_DIR = directory containing this SKILL.md file
Script = {SKILL_DIR}/scripts/gpt-image-gen.sh

For example, if this SKILL.md is at ~/.claude/skills/gpt-image-2-gen/SKILL.md, then the script is at ~/.claude/skills/gpt-image-2-gen/scripts/gpt-image-gen.sh.

After Installation

When this skill is first loaded, proactively greet the user and start the setup:

  1. Check if EVOLINK_API_KEY is set (run: echo $EVOLINK_API_KEY)

    • If not set: "To generate images, you'll need an EvoLink API key. It takes 30 seconds to get one — just sign up at evolink.ai/signup. Want me to walk you through it?"
    • If already set: "You're all set! What kind of image would you like to create?"

  2. That's it. One question. The user is now in the flow.

Do NOT list features, show a menu, or dump instructions. Just ask one question to move forward.

Core Principles

  1. Guide, don't decide — Present options and let the user decide. Don't make assumptions about their preferences.
  2. Let the user drive the creative vision — If they have an idea, use their words. If they need inspiration, offer suggestions and let them choose or refine.
  3. Smart context awareness — Recognize what the user has already provided and only ask about missing pieces.
  4. Intent first — If the user's intent is unclear, confirm what they want before proceeding.

Flow

Step 1: Check for API Key

If the user hasn't provided an API key or set EVOLINK_API_KEY:

  • Tell them they need an EvoLink API Key
  • Guide them to register at https://evolink.ai and get a key from the dashboard
  • Once they provide a key, proceed to Step 2

If the key is already set or provided, skip directly to Step 2.

Step 2: Understand Intent

Assess what the user wants based on their message:

  • Intent is clear (e.g., "generate an image of a sunset") -> Go to Step 3
  • Intent is ambiguous (e.g., "I want to try GPT Image 2") -> Ask what they'd like to do: generate a new image, edit an existing image, learn about model capabilities, etc.

Step 3: Gather Missing Information

Check what the user has already provided and only ask about what's missing:

Parameter What to tell the user Required?
Mode / intent Two modes available: (1) Text-to-image — describe what you want, get an image; (2) Image editing — provide reference image(s) and describe the edits. Determine which mode fits from context, or ask if unclear. Yes
Image content (prompt) Ask what they'd like to see. If they need inspiration, suggest a few ideas for them to pick from or build on. Max 32,000 characters. Yes
Size Supports ratio format (1:1, 16:9, 9:16, etc.) or exact pixels (e.g. 1024x1024). Default: auto (model decides). Only mention if relevant. Optional
Resolution When using ratio format: 1K (~1MP), 2K (~4MP), 4K (~8.3MP). Default: 1K. Only mention for high-quality needs. Optional
Quality low (fast, cheap), medium (balanced, default), high (best, ~4x cost). Only mention if relevant. Optional
Count How many images to generate (1-10). Default: 1. Only ask if the user implies wanting multiple variants. Optional
Reference images For image editing: 1-16 images (JPEG/PNG/WebP, <=50MB each). Ask for URLs if editing mode is detected. Conditional

Smart gathering rules — STRICT:

  • Ask ALL missing parameters in ONE single message. Never split into multiple rounds of questions.
  • Never ask the same question twice. If the user already answered a parameter, it is final — do not re-ask it.
  • Offer defaults upfront so users can say "default is fine": 1:1 / 1K / medium quality / 1 image. If the user says "default" or "just go", use these values immediately.
  • User gives everything at once -> Confirm and generate immediately, no further questions.
  • User gives partial info -> Ask only the remaining missing required fields, all in one message.
  • If user provides image URLs, auto-detect editing mode — no need to ask explicitly.
  • Size, Resolution, Quality, Count all have sensible defaults — if the user skips them, use defaults and proceed.

Step 4: Generate

Once all required information is confirmed:

  1. Before running the script, immediately tell the user: "Got it! Starting your image now — this usually takes 15-60 seconds. I'll keep you posted while it generates."
  2. Run the generation script once. NEVER run it a second time unless the user explicitly asks to retry.
  3. When complete, share the image URL(s) (valid for 24 hours) and generation time.

Execution Pattern by Agent

Claude Code — use the Bash tool with run_in_background: true, then read the output file to check progress:

# Step 1: Run in background (resolves SKILL_DIR first)
Bash(command: "{SKILL_DIR}/scripts/gpt-image-gen.sh \"prompt\" --size 1:1 --quality medium", run_in_background: true, timeout: 300000)

# Step 2: Wait ~30 seconds, then read the background task output file to check for IMAGE_URL= or STATUS_UPDATE lines

# Step 3: When you see IMAGE_URL=<url>, relay it to the user

OpenClaw / OpenCode / Cursor — run the script as a blocking shell command (the script handles its own polling internally, typically completes in 15-120 seconds):

EVOLINK_API_KEY=$EVOLINK_API_KEY {SKILL_DIR}/scripts/gpt-image-gen.sh "prompt" --size 1:1 --quality medium

All agents — critical rules:

  • Once you see TASK_SUBMITTED: in the output, the task is already queued on the server. Do NOT run the script again — retrying wastes the user's API credits.
  • The script blocks until the image is ready (up to 5 minutes). Do NOT kill it prematurely.
  • If the script times out with POLL_TIMEOUT, the image may still be processing — tell the user to check https://evolink.ai/dashboard.

Script Usage

Remember: Replace {SKILL_DIR} with the actual directory containing this SKILL.md file.

# Set API key
export EVOLINK_API_KEY=your_key_here

# Text-to-image (basic)
{SKILL_DIR}/scripts/gpt-image-gen.sh "A beautiful sunset over the ocean"

# Text-to-image with high quality and 4K resolution
{SKILL_DIR}/scripts/gpt-image-gen.sh "Cinematic wide-angle shot of a futuristic city skyline at dusk" --size 16:9 --resolution 4K --quality high

# Text-to-image with exact pixel size
{SKILL_DIR}/scripts/gpt-image-gen.sh "Minimalist logo design" --size 1024x1024 --quality medium

# Image editing (add elements to existing image)
{SKILL_DIR}/scripts/gpt-image-gen.sh "Add a cute kitten next to her" --image "https://example.com/input.png" --size 1:1

# Batch generation (multiple variants)
{SKILL_DIR}/scripts/gpt-image-gen.sh "Pixel art cute robot" --size 1:1 --resolution 2K --quality high --count 4

# With callback URL
{SKILL_DIR}/scripts/gpt-image-gen.sh "Abstract landscape painting" --callback "https://your-domain.com/webhook/image-done"

Script Output Protocol

The script writes structured lines to stdout that you must parse and act on:

Line format When Your action
TASK_SUBMITTED: task_id=<id> estimated=<Ns> Right after submission Confirm to the user that generation has started. This means the API call succeeded — do NOT retry.
STATUS_UPDATE: <message> Every ~15s during generation Relay to the user — e.g., "Still working on your image, about 30 seconds remaining..."
IMAGE_URL=<url> On success (one line per image) Extract the URL and present the image to the user
ELAPSED=<Ns> On success Optionally mention how long it took
POLL_TIMEOUT: task_id=<id> dashboard=<url> Polling exceeded 5 min Tell user: "Your image may already be done — check <dashboard_url> (task: <id>)"
ERROR: ... (stderr) On failure Surface the error message to the user

Critical: Once you see TASK_SUBMITTED:, the task is queued on the server. Do NOT run the script again. Retrying wastes the user's API credits. If the script times out locally, the image may still complete — tell the user to check their dashboard at https://evolink.ai/dashboard.

Error Handling

Provide friendly, actionable messages:

Error What to tell the user
Invalid/missing key (401) "Your API key doesn't seem to work. You can check it at https://evolink.ai/dashboard"
Insufficient balance (402) "Your account balance is low. You can add credits at https://evolink.ai/dashboard"
Rate limited (429) "Too many requests — let's wait a moment and try again"
Content blocked (400) "This prompt was flagged by content moderation. Try adjusting the description"
Image file too large (400) "One of the images is too large. Each image must be <=50MB"
Service unavailable (503) "The service is temporarily busy. Let's try again in a minute"

Model Capabilities Summary

Use this when the user asks what the model can do:

  • Text-to-image: Describe what you want, get an image. Supports prompts up to 32,000 characters.
  • Image editing: Provide 1-16 reference images and describe the edits you want.
  • Sizes: 15 ratio presets (1:1, 16:9, 9:16, 2:1, 3:2, 4:3, 5:4, etc.) + exact pixel format (WxH, 16-aligned, 16-3840px per side)
  • Resolution: 1K (~1MP), 2K (~4MP), 4K (~8.3MP) — only with ratio sizes
  • Quality: low (fast/cheap), medium (balanced), high (best/4x cost)
  • Batch: Generate 1-10 images per request
  • Output: Images valid for 24 hours, save promptly

References

  • references/api-params.md: Complete API parameter reference
  • scripts/gpt-image-gen.sh: Generation script with automatic polling and error handling
Installs
1
Repository
evolinkai/seeda…openclaw
First Seen
10 days ago
Security Audits
Gen Agent Trust HubPass
SnykWarn