Impossible-Studio-TWImpossible-Studio-TW/claude-gpt-image2-mcp/skills/gpt-image/SKILL.md

gpt-image

Skill0 starsChanged 7 months agoMITView on GitHub

Generate images from text using OpenAI's gpt-image-2 (or gpt-image-1 fallback). Triggers when the user types "/gpt-image", says "用 GPT 生圖 / 畫一張", "生圖", "GPT2 / GPT1 畫", "/gpt-image-2 ...", or describes an image they want generated. gpt-image-2 strengths: dense text rendering (great for IG covers, posters with text), photorealism, product photography with accurate text on labels. Default model gpt-image-2 requires OpenAI org verification; falls back to gpt-image-1 (no verification needed) on permission errors.

Reads credentials

19 other projects use a version of this.

Adopted by

Most recently taken up by

potofo/gpt-imagetheir own version, first seen Nov 8, 2026
jujijigo/agent-skillstheir own version, first seen Nov 8, 2026
konglingbukong/gpt-image-studiotheir own version, first seen Oct 20, 2026
AiAi-Mastermind/aiai-mastermind-tools-and-skillstheir own version, first seen Oct 13, 2026
Suge8/friedbuntheir own version, first seen Oct 11, 2026
wuyoscar/gpt_image_2_skilltheir own version, first seen Oct 8, 2026

Other versions19 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 19
wuyoscar/gpt_image_2_skill★ 6ksame name
wuyoscar/GPT-Image2-Skill★ 5.6ksame name
GENEXIS-AI/gpt-image-skill★ 157same name
nikships/droidproxy★ 128same name
Yuki001/game-dev-skills★ 110same name
dshark3y/gpt-image-2-skill★ 19same name
Suge8/friedbun★ 12same name
ZeyuLing/research_autocode_skills★ 9same name
egoist/gpt-image★ 7same name
tuoxie2046/claude-code-research-skills★ 5same name
9 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repositoryOne other file it gives its agents.
- gpt-image-editSkill

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

0
Common
Most of the 15 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
When the user wants Claude to generate an image, call mcp__gpt-image2__gpt_image_generate.
line 8
Very specific:
If the user does NOT specify model / size / quality, ask first with options before calling — the price difference between low and high is …
line 31

“Common” is measured: kept by at least 6 in 10 of the 15 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

/gpt-image — generate images with OpenAI gpt-image-2 / gpt-image-1

When the user wants Claude to generate an image, call mcp__gpt-image2__gpt_image_generate.

Invocation patterns

/gpt-image 一個橘貓在月球上 → call with that prompt
"幫我生一張 IG 直幅封面，內容是 X" → call with size="1024x1536"
"用 GPT2 畫 X" → model="gpt-image-2"
"用 GPT1 / 1 代 / 便宜的畫 X" → model="gpt-image-1"
"高品質 / 漂亮 / 精緻" → quality="high"
"草稿 / 便宜" → quality="low"

Parameters cheat sheet

Arg	Default	Options
`prompt`	(required)	English works best; Chinese OK; include text content verbatim if you want text rendered
`model`	`gpt-image-2`	`gpt-image-2` (newest, needs org verify), `gpt-image-1` (fallback, no verify)
`size`	`1024x1024`	`1024x1024` square / `1024x1536` portrait / `1536x1024` landscape; gpt-image-2 supports custom multiples of 16 up to 3840
`quality`	`auto`	`low` (~~$0.006), `medium` (~~$0.04), `high` (~$0.165), `auto` (model picks)
`save_to`	(auto)	Absolute path; defaults to `~/Pictures/gpt-image2/` or workspace folder

Asking the user before defaulting (cost-sensitive)

If the user does NOT specify model / size / quality, ask first with options before calling — the price difference between low and high is ~28×. Suggested format:

要哪一組？
(A) 草稿 — gpt-image-1 + low + 方形 — ~$0.006
(B) 標準 — gpt-image-1 + medium + 方形 — ~$0.04
(C) 最強 — gpt-image-2 + high + 直幅 — ~$0.165 (requires org verify)

Skip asking if user said "draft / 隨便 / 測試" → use cheapest defaults.

Returning results

Tool returns the absolute path to the saved PNG. Display that path to the user. If you can attach files, attach the PNG. If the response starts with [, it's an error — explain in plain language and suggest a fix (often: install missing deps, set OPENAI_API_KEY in .env, or org verify).

Common errors

Error	Cause	Fix
`[Error: PermissionDeniedError 403 ... organization must be verified]`	gpt-image-2 needs org verify	Verify at https://platform.openai.com/settings/organization/general OR use `model="gpt-image-1"`
`[Error: AuthenticationError]`	API key missing or invalid	Check `.env` in plugin dir has `OPENAI_API_KEY=...`
`[Error: RateLimitError]`	Hit OpenAI quota	Wait, or use a different account
`[Error: BadRequestError]`	Prompt violates content policy	Rephrase; no celebrities/copyrighted characters/violence
MCP timeout (60s)	High quality + large size takes 60-90s	Image still saves to disk — read from `save_to` location

Tips for good prompts

Be specific: subject, action, environment, lighting, mood, framing, style
For text rendering (gpt-image-2 specialty): put the text content in quotes verbatim, e.g. ... a poster with the text "週末特賣 — 全店 5 折"
For consistency: identity preservation across prompts is weak in pure generate mode; use gpt_image_edit with reference images for character work

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free