Skip to content
Agentmart
Submit a tool

gpt-image

Generate and edit images with OpenAI GPT Image 2.5 via fal.ai.

Skill by sejas

IndexedMIT★ 2
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/sejas-gpt-image

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Generate and edit images with OpenAI GPT Image 2.5 via fal.ai. Use when the user wants an image from a text prompt, or wants an existing image edited while keeping subject, composition and background intact. Covers all four endpoints — flare (fast default) and sunburst (premium detail / tightest edit control), each in text-to-image and edit form. Supports transparent backgrounds, preset or custom sizes (e.g. 1792×1024 for 16:9 slides), quality up to max, up to 10 images per request, masks, and up to 16 reference images. Stdlib-only Python; no pip install required.

Use cases: presentations, writing, agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 26, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Python package

Network hosts it talks to 2

  • codeload.github.com · found in install command
  • queue.fal.run · found in code

Files it touches

  • ~/.claude/skills/gpt-image
  • Your home directory (~)

What's inside

Skill name
gpt-image
Description
Generate and edit images with OpenAI GPT Image 2.5 via fal.ai. Use when the user wants an image from a text prompt, or wants an existing image edited while keeping subject, composition and background intact. Covers all four endpoints — flare (fast default) and sunburst (premium detail / tightest edit control), each in text-to-image and edit form. Supports transparent backgrounds, preset or custom sizes (e.g. 1792×1024 for 16:9 slides), quality up to max, up to 10 images per request, masks, and up to 16 reference images. Stdlib-only Python; no pip install required.
License
MIT
Instructions
185 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts
  • Instructions call the network
  • Instructions read files in your home folder (~/.ssh, ~/.aws…)

Files (2)

  • SKILL.md
  • generate.py · scripts/

SKILL.md (first lines)

# GPT Image 2.5 (fal.ai)

One script drives all four `openai/gpt-image-2.5` endpoints on fal.ai's queue API.

| Endpoint | Reach it with | Best for |
|---|---|---|
| `flare/text-to-image` | *(default)* | **Start here for any new image.** Fast, high quality. Natural lighting, rich textures, complex layouts, transparent backgrounds. |
| `flare/edit` | `--image` | **Start here for any edit.** Changes only what's asked; subject, composition and background survive. References stay recognisable across styles and successive edits. |
| `sunburst/text-to-image` | `--tier sunburst` | Premium hero images. Extra fidelity on intricate detail — fine typography, ornament, texture, product shots. Slower; pay the wait only when detail is the point. |
| `sunburst/edit` | `--tier sunburst --image` | Tightest control. Edits scoped precisely to the instruction, subject and composition preserved across many revision rounds. Reach for it when `flare/edit` drifts. |

Rule of thumb: **flare until it fails you, then sunburst.** Flare is the default for most
work; sunburst buys detail and obedience at the cost of time.

### Choosing between this and the other image skills

- **This skill** — OpenAI-style prompting, transparent backgrounds, masked edits, strong
  in-image typography, sizes pinned to exact pixel dimensions.
- **`fal-image-edit`** (nano-banana-pro / Gemini 3 Pro on fal) — better when composing
  **many references** (up to 14), when you need **web-search grounding** for factual
  content, **4K** output, a **seed** for reproducibility, or aspect-ratio-by-name control.
- **`nano-banana-image-editor`** — same model family through a direct Google key rather
  than fal billing.

## Prerequisites

A fal.ai API key. Set once:

```bash
export FAL_KEY=your_key_here
# or persist:
mkdir -p ~/.config/fal && echo your_key_here > ~/.config/fal/key
```

Get a key at https://fal.ai/dashboard/keys.

## Usage

Mode is implicit: **no `--image` means text-to-image, `--image` means edit.**

```bash
SKILL=${CLAUDE_PLUGIN_ROOT:-$HOME/.claude}/skills/gpt-image/scripts/generate.py

# Generate (flare, the default)
python3 $SKILL \
  --prompt "A serene minimalist illustration of a lighthouse at sunrise, warm cream palette" \
  --output ~/out.png
```

### Premium detail

```bash
python3 $SKILL --tier sunburst --quality xhigh \
  --prompt "Art-nouveau poster, intricate gold filigree border, letterpress texture" \
  --output ~/hero.png
```

### Edit an existing image

`--image` takes a **local path or a URL**. Local files upload to fal storage automatically.

```bash
python3 $SKILL \
  --image ~/photo.jpg \
  --prompt "Make it night-time, add neon signs in the background" \
  --output ~/out.png
```

### Edit with tightest control, across revision rounds

```bash
python3 $SKILL --tier sunburst \
  --image ~/poster.png \
  --prompt "Replace only the date line with '11-13 December 2026'. Leave everything else untouched." \
  --output ~/poster-v2.png
```

### Masked edit — confine changes to a region

```bash
python3 $SKILL --image ~/room.png --mask ~/room-mask.png \
  --prompt "Put a large window here" --output ~/room-window.png
```

### Transparent background (logos, stickers, overlays)

```bash
python3 $SKILL --background transparent --output-format png \
  --prompt "A flat vector capoeira berimbau icon, single colour, no background" \
  --output ~/icon.png
```

### 16:9 for presentation slides

```bash
python3 $SKILL --size 1792x1024 --quality high \
  --prompt "..." --output ~/slide.png
```

### Batch variations

```bash
python3 $SKILL --num-images 4 --prompt "..." --output ~/out.png
# saves ~/out-1.png … ~/out-4.png
```

### All flags

| Flag | Default | Notes |
|---|---|---|
| `--prompt` | required | Prompt, or the editing instruction |
| `--output` | required | Local path; parent dirs auto-created |
| `--tier` | `flare` | `flare` or `sunburst` |
| `--image` | — | Local path or URL. Repeatable, up to **16**. Presence switches to edit mode |
| `--mask` | — | Edit mode only; confines changes to a region |
| `--size` | `landscape_4_3` generating, `auto` editing | Preset, `auto`, or `WxH` |
| `--quality` | `high` | `auto`, `low`, `medium`, `high`, `xhigh`, `max` |
| `--background` | `auto` | `auto`, `transparent`, `opaque` |
| `--num-images` | `1` | **1–10**. >1 suffixes filenames `-1`, `-2`, … |

From sejas/ai-code-skills, MIT.

More from sejas/ai-code-skills

Signals

Stars
2
Forks
1
Open issues
0
Last commit (this folder)
Oct 8, 2026 (20 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from sejas

Cite this page

gpt-image skill. Agent Mart. https://agentmart.dev/skills/sejas-gpt-image (last updated 2026-10-24)