imagegen

name	imagegen
description	Use when the user asks to generate or edit images (e.g., generate image, edit/inpaint/mask, background removal, transparent background, product shots, concept art, covers, or batch variants). Runs built-in image_generation_call first, then falls back to custom endpoint via auth.json.

Image Generation Skill

Generates or edits images for the current project (website assets, game assets, UI mockups, product mockups, logos, photorealistic images, infographics).

When to use

• Generate a new image (concept art, product shot, cover, website hero)
• Edit an existing image (inpainting, masked edits, background replacement, object removal)
• Batch runs (many prompts or many variants)

Decision tree (generate vs edit vs batch)

• If the user provides an input image OR says "edit/retouch/inpaint/mask" → edit
• Else if the user needs many different prompts/assets → generate-batch
• Else → generate

---

Environment — Priority Order

1st choice: Built-in image_generation_call (PREFERRED)

If built-in tool is available in this session → use it directly.
Best quality, no API key needed, platform-handled.
Skip all steps below if this works.

2nd choice: Custom endpoint via auth.json

Only if built-in tool is NOT available in this session.

Read credentials from ~/.config/imagegen/auth.json:

{
  "endpoint": "https://your-proxy-or-api/v1/images/generations",
  "api_key": "YOUR_KEY_HERE",
  "model": "cx/gpt-5.4-image"
}

Auto-detect and run:

python3 - <<'PY'
import json, os, subprocess, sys
from pathlib import Path

auth_path = Path.home() / '.config' / 'imagegen' / 'auth.json'
if not auth_path.exists():
    sys.exit("[imagegen] auth.json not found. See SETUP.md for instructions.")

auth = json.loads(auth_path.read_text())
endpoint = auth.get('endpoint', '').strip()
api_key  = auth.get('api_key', '').strip()
model    = auth.get('model', 'gpt-image-1').strip()

if not endpoint or not api_key:
    sys.exit("[imagegen] auth.json is missing 'endpoint' or 'api_key'.")

print(f"[imagegen] Using endpoint: {endpoint}")
print(f"[imagegen] Model: {model}")
PY

Full curl call:

python3 scripts/image_gen.py generate \
  --prompt "<PROMPT>" \
  --model "$(python3 -c "import json,pathlib; d=json.loads(pathlib.Path.home().joinpath('.config/imagegen/auth.json').read_text()); print(d.get('model','gpt-image-1'))")"\
  --out output/imagegen/<filename>.png

OR via direct curl (SSE stream):

ENDPOINT=$(python3 -c "import json,pathlib; d=json.loads(pathlib.Path.home().joinpath('.config/imagegen/auth.json').read_text()); print(d['endpoint'])")
API_KEY=$(python3 -c "import json,pathlib; d=json.loads(pathlib.Path.home().joinpath('.config/imagegen/auth.json').read_text()); print(d['api_key'])")
MODEL=$(python3 -c "import json,pathlib; d=json.loads(pathlib.Path.home().joinpath('.config/imagegen/auth.json').read_text()); print(d.get('model','gpt-image-1'))")

curl -sS -N -X POST "$ENDPOINT" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Accept: text/event-stream" \
  -d "{\"model\":\"$MODEL\",\"prompt\":\"<PROMPT>\",\"n\":1,\"size\":\"auto\",\"quality\":\"high\",\"output_format\":\"png\"}" \
  -o tmp/imagegen/response.raw

# Parse SSE and save PNG
python3 - <<'PY'
import json, base64
from pathlib import Path

raw = Path('tmp/imagegen/response.raw').read_text(errors='replace')
final = None
for block in raw.strip().split('\n\n'):
    lines = block.splitlines()
    ev   = next((l[6:].strip() for l in lines if l.startswith('event:')), '')
    data = ''.join(l[5:].strip() for l in lines if l.startswith('data:'))
    if ev == 'done' and data:
        final = json.loads(data)

if not final:
    raise SystemExit('[imagegen] No done event in SSE response.')

out = Path('output/imagegen/result.png')
out.parent.mkdir(parents=True, exist_ok=True)
out.write_bytes(base64.b64decode(final['data'][0]['b64_json']))
print(f)
PY

STOP conditions (strict)

• Do NOT hardcode any API key or endpoint URL in code or skill files
• Do NOT use direct OpenAI API (api.openai.com) unless user explicitly sets it in auth.json
• Do NOT use sharp, canvas, jimp, PIL, or any programmatic library as image substitute
• Do NOT fake image generation with placeholder or solid colour
• If built-in tool unavailable AND auth.json missing → inform user to create auth.json, then STOP
• If both option 1 and 2 fail → report clearly and STOP

---

Workflow

0. Environment check (strict order):

   • a. Built-in image_generation_call tool available? → use it, proceed to step 1
   • b. ~/.config/imagegen/auth.json exists and valid? → use curl/SDK path, proceed
   • c. Neither available → STOP. Report: "Built-in tool unavailable and auth.json not found. See SETUP.md."
   • d. NEVER ask for or use a hardcoded API key

1. Decide intent: generate vs edit vs batch (see decision tree above)

2. Collect inputs: prompt(s), exact text (verbatim), constraints, and any input image(s)/mask(s)

3. If batch: write temp JSONL under tmp/imagegen/, run once, delete JSONL after

4. Augment prompt into a short labeled spec (structure + constraints)

5. Run image generation via priority order above

6. Inspect outputs and validate: subject, style, composition, text accuracy, constraints

7. Iterate with single targeted change if needed

8. Save final outputs to output/imagegen/ and note the final prompt used

---

Defaults & rules

• Default model from auth.json → field model. If missing, default to gpt-image-1
• Assume the user wants a new image unless they explicitly ask for an edit
• Prefer built-in tool over curl path always
• Prefer the bundled CLI (scripts/image_gen.py) over writing new one-off scripts
• Never modify scripts/image_gen.py
• Use tmp/imagegen/ for intermediates, output/imagegen/ for final assets

---

Prompt augmentation

Reformat user prompts into a structured spec. Only make implicit details explicit; do not invent new requirements.

Template:

Use case: <taxonomy slug>
Asset type: <where the asset will be used>
Primary request: <user's main prompt>
Scene/background: <environment>
Subject: <main subject>
Style/medium: <photo/illustration/3D/etc>
Composition/framing: <wide/close/top-down; placement>
Lighting/mood: <lighting + mood>
Color palette: <palette notes>
Quality: <low/medium/high/auto>
Text (verbatim): "<exact text>"
Constraints: <must keep/must avoid>
Avoid: <negative constraints>

Use-case taxonomy

Generate: photorealistic-natural, product-mockup, ui-mockup, infographic-diagram, logo-brand, illustration-story, stylized-concept, historical-scene
Edit: text-localization, identity-preserve, precise-object-edit, lighting-weather, background-extraction, style-transfer, compositing, sketch-to-render

---

Reference map

• references/cli.md — how to run image_gen.py CLI
• references/image-api.md — API parameters reference
• references/prompting.md — prompting principles
• references/sample-prompts.md — copy-paste prompt recipes
• SETUP.md — user setup instructions (auth.json, dependencies)