Nano Banana Pro Prompts Recommend Skill
YouMind-OpenLab/nano-banana-pro-prompts-recommend-skill
Recommend suitable prompts from 10,000+ Nano Banana Pro image generation prompts based on user needs.
This skill should be used when the user asks to "generate an image", "create a logo", "draw an icon", "edit this photo", "change background to transparent", "remove background", "use GPT image"…
$ npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install Wangnov/gpt-image-2-skill gpt-image-2-skill --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/Wangnov/gpt-image-2-skill.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/gpt-image-2-skill .claude/skills/gpt-image-2-skill && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "gpt-image-2-skill" agent skill from https://github.com/Wangnov/gpt-image-2-skill/tree/main/skills/gpt-image-2-skill into .claude/skills/gpt-image-2-skill/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "gpt-image-2-skill", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/Wangnov/gpt-image-2-skill/tree/main/skills/gpt-image-2-skillType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install Wangnov/gpt-image-2-skill gpt-image-2-skill --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Wangnov/gpt-image-2-skill.git skills-src && mkdir -p .agents/skills && cp -r skills-src/skills/gpt-image-2-skill .agents/skills/gpt-image-2-skill && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "gpt-image-2-skill" agent skill from https://github.com/Wangnov/gpt-image-2-skill/tree/main/skills/gpt-image-2-skill into .agents/skills/gpt-image-2-skill/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "gpt-image-2-skill", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install Wangnov/gpt-image-2-skill gpt-image-2-skill --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Wangnov/gpt-image-2-skill.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/skills/gpt-image-2-skill .cursor/skills/gpt-image-2-skill && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "gpt-image-2-skill" agent skill from https://github.com/Wangnov/gpt-image-2-skill/tree/main/skills/gpt-image-2-skill into .cursor/skills/gpt-image-2-skill/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "gpt-image-2-skill", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/Wangnov/gpt-image-2-skill.git --path skills/gpt-image-2-skill--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install Wangnov/gpt-image-2-skill gpt-image-2-skill --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Wangnov/gpt-image-2-skill.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/skills/gpt-image-2-skill .gemini/skills/gpt-image-2-skill && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "gpt-image-2-skill" agent skill from https://github.com/Wangnov/gpt-image-2-skill/tree/main/skills/gpt-image-2-skill into .gemini/skills/gpt-image-2-skill/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "gpt-image-2-skill", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install Wangnov/gpt-image-2-skill gpt-image-2-skillInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/Wangnov/gpt-image-2-skill.git skills-src && mkdir -p .github/skills && cp -r skills-src/skills/gpt-image-2-skill .github/skills/gpt-image-2-skill && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "gpt-image-2-skill" agent skill from https://github.com/Wangnov/gpt-image-2-skill/tree/main/skills/gpt-image-2-skill into .github/skills/gpt-image-2-skill/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "gpt-image-2-skill", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install Wangnov/gpt-image-2-skill gpt-image-2-skill --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Wangnov/gpt-image-2-skill.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/skills/gpt-image-2-skill .opencode/skills/gpt-image-2-skill && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "gpt-image-2-skill" agent skill from https://github.com/Wangnov/gpt-image-2-skill/tree/main/skills/gpt-image-2-skill into .opencode/skills/gpt-image-2-skill/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "gpt-image-2-skill", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
gpt-image-2-skillThis skill should be used when the user asks to "generate an image", "create a logo", "draw an icon", "edit this photo", "change background to transparent", "remove background", "use GPT image"…
Gpt Image 2 Skill is an agent skill from Wangnov/gpt-image-2-skill. This skill should be used when the user asks to "generate an image", "create a logo", "draw an icon", "edit this photo", "change background to transparent", "remove background", "use GPT image", "use Codex to draw", "用 GPT image 生成图片", "用 Codex 画图", "帮我生成一张图", "改成透明背景", "把这张图编辑一下", or any prompt-to-image or reference-image-edit task that benefits from a structured CLI returning JSON results and JSONL progress events. Supports OpenAI GPT Image models (via OPENAIAPIKEY or OpenAI-compatible base URL) and Codex…
Its SKILL.md is about 4.5k tokens, which your agent loads only when the skill is triggered. The skill folder holds 18 other files, including scripts and reference files (for example `agents/openai.yaml`, `evals/README.md` and `evals/evals.json`).
It sits in Media & Creative, covering Image generation. It works with OpenAI and Tauri. The repository describes itself as: GPT Image 2 Skill / CLI / Web / Desktop. The licence is MIT.
Read from SKILL.md and the folder at commit 16fcc05. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Ships 2 files in scripts/ (JavaScript), which the agent can run.
Shell commands in SKILL.md call:
nodenpmcargoFrom the folder's file list and the shell code blocks in SKILL.md.
Links to these hosts (documentation or services it may open):
developers.openai.comFrom URLs in SKILL.md, links to its own repository left out.
Names these keys or tokens, usually read from environment variables:
OPENAI_API_KEYFrom names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Gpt Image 2 Skill loads about 4.5k tokens when it runs, and up to ~18k if it reads all its reference files. Until then it costs about 187 tokens; SKILL.md has 1,549 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.
The full file from Wangnov/gpt-image-2-skill at commit 16fcc05, republished under its MIT licence (© Wangnov). 1,549 words, ~4,487 tokens.
.claude/skills/gpt-image-2-skill/SKILL.md (or your agent's skills folder). This skill also uses 14 other files; get the full folder from GitHub.Run image generation and editing through one CLI surface that hides provider differences. The Node wrapper at scripts/gpt_image_2_skill.cjs resolves an underlying Rust binary (env override → bundled Skill binary → installed binary → Tauri App bundled CLI → repo cargo run → cached release → bootstrap download) and forwards every flag. On glibc Linux, release bootstrap tries the GNU archive first and then the static musl archive as the sandbox fallback.
OPENAI_API_KEY, an OpenAI-compatible base URL, and Codex auth.json without changing command shape.$CODEX_HOME/gpt-image-2-skill/config.json so CLI, App, and Skill use the same default provider.Always pass --json so the result is machine-readable. Add --json-events when progress visibility matters.
# 1. Confirm runtime + provider readiness
node scripts/gpt_image_2_skill.cjs --json config inspect
node scripts/gpt_image_2_skill.cjs --json doctor
node scripts/gpt_image_2_skill.cjs --json auth inspect
# 2. Generate a final transparent PNG deliverable
node scripts/gpt_image_2_skill.cjs --json --json-events \
transparent generate --prompt "..." --out /tmp/asset.png \
--size 2K --quality high
# 3. Generate a normal image (auto-selects provider; OpenAI first, then Codex)
node scripts/gpt_image_2_skill.cjs --json --json-events \
images generate --prompt "..." --out /tmp/out.png \
--format png --size 2K
# 4. Edit a reference image (OpenAI multipart)
node scripts/gpt_image_2_skill.cjs --json --json-events \
images edit --prompt "..." --ref-image /tmp/in.png --out /tmp/out.png
# 5. Remove a controlled background from existing source images
node scripts/gpt_image_2_skill.cjs --json \
transparent extract --input /tmp/source-green.png --out /tmp/asset.png \
--method chroma --matte-color auto --strict
# 6. Verify the final file before delivery
node scripts/gpt_image_2_skill.cjs --json \
transparent verify --input /tmp/asset.png --profile icon --strict
# 7. Raw request escape hatch
node scripts/gpt_image_2_skill.cjs --json \
request create --request-operation generate \
--body-file /tmp/body.json --out-image /tmp/out.png --expect-image
# 8. Self-test (calls doctor + auth inspect)
node scripts/selftest.cjsForce a provider with --provider openai, --provider codex, or any named provider from config inspect; leave the default --provider auto to use default_provider first. Override the legacy OpenAI base URL with --openai-api-base https://....
以下 CLI 示例走 Responses 路径。复现最新 Codex App 时,应区分宿主原生 image_gen.imagegen 的独立 Images 路径;见 references/codex-native-imagegen.md。原生工具成功不代表 CLI 已适配,服务端别名也不能证明实际权重版本。
用户指定 Codex 时,必须显式传 --provider codex,避免共享配置将请求路由到 OpenAI-compatible 服务。下面的命令从 Skill 目录执行,使用本地 Codex 登录,不需要 OpenAI API key:
node scripts/gpt_image_2_skill.cjs --json --provider codex doctor
mkdir -p output
node scripts/gpt_image_2_skill.cjs --json --json-events --provider codex \
images generate --model gpt-6-astra \
--prompt "一只橙色纸雕小狐狸坐在蓝绿色月牙上,深蓝背景" \
--out output/codex-fox.png --format png --size 1024x1024 \
> output/result.json 2> output/events.jsonlgpt-5.4;本地实测已被 ChatGPT 账户后端拒绝。示例显式使用本次验证的 gpt-6-astra,其他账户仍需核对可用模型。--model 是外层调用工具的模型,不能填 gpt-image-2.5-flare 或 gpt-image-2.5-sunburst。这两个是图像模型 ID。tools[].model,由 Codex 服务端选择绘图模型。读取 类型为 response.created 的 SSE 日志中 data.response.tools[].model;不要将 CLI 摘要的 delegated_image_model 当作实测结果,该字段在当前版本仍硬编码为 gpt-image-2。ok: true、图片文件真实存在且经过视觉检查,才能报告生图成功。请求被接受或进入 generating 不能视为完成。references/codex-local-verification.md。OpenAI 官方已列出 gpt-image-2.5-flare 和 gpt-image-2.5-sunburst,可在 OpenAI Image API 中使用 --model 显式选择;这不代表 Codex 订阅后端已支持相同模型选择。保留用户指定的模型和共享配置。
2.5 的公共 API 支持 xhigh / max 质量和原生透明 PNG/WebP,当前 Rust CLI 与桌面生成界面已支持 low|medium|high|xhigh|max|auto;xhigh/max 仅用于支持它们的 API 模型,不能保证 Codex 订阅后端支持。Codex 的原始请求操作必须是 responses,不能用 generate。
Before using newly documented command groups, especially transparent generate, transparent extract, or transparent verify, confirm the actual CLI runtime is fresh enough. Skill files, the Tauri App bundle, npm global CLI, cargo-installed CLI, and wrapper cache can drift independently.
Start by checking the latest upstream CLI version:
npm view gpt-image-2-skill versionThen look for the Tauri App bundled CLI first. Treat the app bundle as user-managed software: if it exists but is older than upstream, ask the user before updating or replacing the app.
APP_CLI="/Applications/GPT Image 2.app/Contents/Resources/bin/gpt-image-2-skill"
[ -x "$APP_CLI" ] && "$APP_CLI" --versionIf no app bundle CLI exists, check the shell CLI. If it is missing or older than upstream, silently update the CLI dependency with npm, then verify that the current shell resolves the new binary:
command -v gpt-image-2-skill || true
gpt-image-2-skill --version || true
npm install -g gpt-image-2-skill@latest
command -v gpt-image-2-skill
gpt-image-2-skill --versionAfter any update, run the skill wrapper smoke checks:
node scripts/gpt_image_2_skill.cjs --json doctor
node scripts/selftest.cjsIf a documented subcommand fails with unrecognized subcommand, diagnose stale runtime first rather than changing prompts or providers. Prefer the wrapper in this skill directory for reproducible skill execution, but keep the bare CLI fresh when examples or user commands call gpt-image-2-skill directly.
Use the CLI config surface when the user asks to add or pin a provider:
node scripts/gpt_image_2_skill.cjs --json config path
node scripts/gpt_image_2_skill.cjs --json config add-provider \
--name my-image-api \
--type openai-compatible \
--api-base https://example.com/v1 \
--api-key sk-... \
--set-default
node scripts/gpt_image_2_skill.cjs --json config test-provider my-image-apiCredential sources supported by CLI, App, and Skill: file, env, and keychain. File credentials are stored in the shared config file; JSON output redacts them.
输出属性优先通过接口参数表达,但 Codex 后端可能归一化或忽略参数。提示词可表达期望,最终必须读取实际返回值并检查图片;不要将任何单一参数或提示词视为执行保证。
| Property | CLI flag |
|---|---|
| Output background (transparent / opaque / auto) | --background auto|transparent|opaque |
| Output dimensions | --size 2K, --size 4K, or --size WIDTHxHEIGHT |
| Output container | --format png|jpeg|webp |
| Compression level | --compression 0..100 |
| Render quality | --quality low|medium|high|auto |
| Number of images | --n <count> (OpenAI only) |
| Edit mask region | --mask <png> (OpenAI only) |
提示词用于内容与视觉意图;--background opaque 只表达不透明,不指定白色。白背景还需在提示词中明确。对于 Codex,透明提示词配合 background=auto 可能得到真实 alpha,而显式透明参数当前可能报错;以实测和文件检查为准。
Provider asymmetry: --n, --moderation, --mask 和 --input-fidelity 在 Codex 普通命令中返回 unsupported_option。--background 会写入 Codex 请求,但不能据此保证上游执行;本次原始透明背景请求被后端拒绝。公共 Image API 与 Codex 后端的能力必须分别验证。
Codex 的显式 background=transparent 在本地对照中被拒绝,但 background=auto 配合透明提示词曾成功生成真实 RGBA。可以将后者作为候选路径,必须验证 alpha 和严格交付指标;本次样本因全透明区域残留 RGB 未通过严格验收。已合格的原生透明图无需再次抠图。需要本地处理时使用下列流程;OpenAI 2.5 公共 API 的原生透明输出也需同样验收:
transparent generate — prompt-to-final PNG. It generates a controlled matte source, extracts alpha locally, verifies the result, and only succeeds when the final PNG passes transparency checks.transparent extract — local background removal from controlled source images you generated yourself. It is not a general-purpose background remover for arbitrary photos.transparent verify — final gate for any PNG before delivery. Use --strict and the right --profile when the file must be accepted or fail the task.A transparent deliverable is valid only if the final file has a real PNG alpha channel and passes verification. A visual appearance of transparency, a white background, or a checkerboard pattern is not sufficient.
--strict is profile-based:
| Profile | Use for | Extra strictness |
|---|---|---|
generic | common alpha/file checks | does not over-police unusual assets |
icon | clean single-subject icons and props | requires clean opaque core, margin, low stray noise |
product | product/object cutouts | similar to icon, with residue and edge checks |
sticker | decals, badges, multi-detail props | allows more intentional small components than icon |
seal | stamps, seals, logos with inner marks | allows split components such as ring + center symbol |
translucent | glass, liquid, crystal | requires partial alpha |
glow | light ribbons, flame, smoke, particles | requires partial alpha and transparent margin |
shadow | soft shadow assets | requires partial alpha and transparent margin |
effect | hard-alpha particles, bursts, UI effects | transparent margin without requiring partial alpha |
The CLI is intentionally not a material classifier. The Agent should choose generation prompts and extraction methods based on the asset:
| Asset type | Generation guidance | Extraction guidance |
|---|---|---|
| Opaque object, icon, sticker, product | Single isolated subject, clear margin, perfectly flat chroma matte. Pick a matte color absent from the object. | transparent generate or transparent extract --method chroma --matte-color auto |
| Thin edges, hair, fur, lace, chain, netting | Use high resolution, strong subject/background contrast, no contact shadow, no background-colored details. Try magenta/cyan/green mattes if one contaminates the edge. | Chroma extraction with --spill-suppression when needed, then verify with --expected-matte-color; retry with a different matte if residue remains. |
| Glass, crystal, liquid, hologram | Ask for a centered asset on flat black and flat white backgrounds, keeping geometry identical. Use reference/edit flow when possible to keep alignment. | transparent extract --method dual --dark-image black.png --light-image white.png |
| Glow, flame, smoke, mist, magic particles | Generate dark and light background variants. Avoid textured backgrounds and avoid bloom reaching the image edge unless the edge is intentional. | Prefer dual extraction; verify that partial_pixels is non-zero. |
| Shadows | Decide whether the shadow is part of the asset. If not, explicitly forbid contact shadows. If yes, generate on a flat matte with enough margin. | Chroma for opaque shadow silhouettes; dual extraction for soft translucent shadows. |
| Unknown or unusual material | Do not classify it first. Generate controlled source variants, run extraction candidates, and keep the one that passes verification with the cleanest edge. | Use --report-dir / --keep-sources while iterating, then deliver only the final PNG. |
For chroma extraction, --matte-color auto samples the actual flat source background from the image edges. Prefer it when the source was AI-generated, because prompts like "pure #ff00ff" often produce near-matte colors rather than exact RGB values. Use explicit --matte-color <name|#rrggbb> only when the source background is known exactly.
For extraction tuning, use --material only as a broad hint, not as a subject classifier: standard, soft-3d, flat-icon, sticker, or glow. Manual --threshold, --softness, and --spill-suppression override the selected preset.
For style-locked transparent assets, transparent generate is prompt-only. Use a flat RGB reference image with images edit --ref-image to create a controlled matte source, then run transparent extract. Do not use a transparent PNG as the reference image unless you intentionally want the alpha/composited edge behavior to influence the edit.
GPT Image 2 can render accurate UI text, numbers, scores, labels, and logo marks in the bitmap when they are part of the desired artwork. Put the exact wording or numbering in the prompt and verify the output visually. Render text separately in the host app or design tool only when it must stay editable, localizable, programmatically changeable, or perfectly consistent across many generated variants.
Examples:
# Simple asset: final transparent PNG, sources hidden unless there is a failure
node scripts/gpt_image_2_skill.cjs --json --json-events \
transparent generate \
--prompt "a polished fantasy sword game asset, no text, no frame" \
--out /tmp/sword.png --size 2K --quality high
# Agent-controlled chroma flow
node scripts/gpt_image_2_skill.cjs --json --json-events \
images generate \
--prompt "a silver necklace, centered, on a perfectly flat pure magenta background, no shadow" \
--out /tmp/necklace-magenta.png --format png --size 2K
node scripts/gpt_image_2_skill.cjs --json \
transparent extract --method chroma \
--input /tmp/necklace-magenta.png --matte-color auto \
--out /tmp/necklace.png --material sticker --strict
# Semi-transparent material flow
node scripts/gpt_image_2_skill.cjs --json \
transparent extract --method dual \
--dark-image /tmp/glow-on-black.png \
--light-image /tmp/glow-on-white.png \
--out /tmp/glow.png --strictAlways inspect the JSON verification fields before delivery: passed, alpha_min, alpha_max, transparent_ratio, partial_pixels, and warnings. Also inspect quality fields: checkerboard_detected, touches_edge, edge_margin_px, stray_pixel_count, largest_component_ratio, matte_residue_checked, matte_residue_score, halo_score, transparent_rgb_scrubbed, alpha_health_score, residue_score, quality_score, and failure_reasons. If passed is false, do not deliver the file as a transparent PNG. If matte_residue_checked is false for a chroma-derived PNG, run transparent verify again with the source matte via --expected-matte-color.
gpt-image-2 / Codex gpt-5.4;Codex 使用上面的显式模型示例,不能依赖旧默认值。--size、--quality、--format、--compression;Codex 服务端可能归一化这些参数,应核对返回值和实际文件。--n, --moderation, --mask, --input-fidelity; Codex background behavior requires separate verification.401 triggers one token refresh + one retry.2K → 2048x2048, 4K → 3840x2160. Custom WxH requires both edges multiples of 16, max edge 3840, max 8,294,400 pixels, max aspect ratio 3:1.Load on demand for deeper detail:
references/codex-native-imagegen.md — App 原生工具、独立 Images 端点与旧 CLI 调用路径的区别。references/codex-local-verification.md — 本地 Codex 生图命令、实测结果与模型识别边界。references/providers.md — OpenAI / OpenAI-compatible / Codex selection, auth sources, runtime discovery, update policy, and resolution order.references/sizes-and-formats.md — size aliases, custom constraints, format/quality/compression/background, shared vs OpenAI-only flags.references/transparent-png.md — Agent playbook for prompt design, controlled mattes, dual-background extraction, verification, and retry loops.references/json-output.md — --json stdout schema, success and error envelopes, per-command shapes.references/json-events.md — --json-events JSONL phases (request_started, multipart_prepared, retry_scheduled) and Codex SSE passthrough.references/troubleshooting.md — runtime_unavailable, auth_missing, Codex 401 refresh, retry policy, size rejections, moderation, timeouts.The companion file agents/openai.yaml is read by Codex Skill runtime only (Claude Code ignores it). Both runtimes execute the commands above with cwd at the skill directory, so relative paths like scripts/gpt_image_2_skill.cjs resolve in either harness.
© Wangnov, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
SKILL.md and 14 other files (scripts, references) in skills/gpt-image-2-skill of Wangnov/gpt-image-2-skill.
Open the folder on GitHubat commit 16fcc05
Gpt Image 2 Skill next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Gpt Image 2 Skill this skillWangnov/gpt-image-2-skill | 142 | — | ~4.5k | Automated safety check: Pass | MIT | |
| Nano Banana Pro Prompts Recommend SkillYouMind-OpenLab/nano-banana-pro-prompts-recommend-skill | 1.9k | 1 repos | ~4.1k | Automated safety check: Pass | None | |
| Yingzaoop7418/guizang-yingzao-skill | 496 | — | ~1.1k | Automated safety check: Pass | None | |
| AI Image Creatorcentminmod/my-claude-code-setup | 2.7k | — | ~8.1k | Automated safety check: Notes | MIT | |
| Character Refseternityspring/shuohao-skills | 4.3k | — | ~1.7k | Automated safety check: Warn | Apache-2.0 | |
| Basic Memory Repo Imagesbasicmachines-co/basic-memory | 4.1k | — | ~2.7k | Automated safety check: Pass | AGPL-3.0 |
YouMind-OpenLab/nano-banana-pro-prompts-recommend-skill
Recommend suitable prompts from 10,000+ Nano Banana Pro image generation prompts based on user needs.
op7418/guizang-yingzao-skill
Transform real Chinese architecture and place-based cultural photos into art-directed editorial posters, integrated multi-photo scenes, and optional source comparisons.
centminmod/my-claude-code-setup
Generate, edit-from-reference, or analyze images with AI via OpenRouter (Gemini, GPT Image, Seedream, Qwen, MAI, Grok, FLUX.2, Recraft, Muse, Riverflow; Cloudflare AI Gateway BYOK).
eternityspring/shuohao-skills
给任何故事里的角色真出参考图(小说改编、自己原创的故事、单独设计一个角色都行,不需要小说原文): 一段话描述角色,拆成分层字段、补全后确认, 先出一张正面全身锚点,其余视图(大头照、90° 侧面、背面、细节、45° 大头照)都只参考这张锚点, 按需分档出图。每张图带标识、可单独重出,重出后自动标出哪些图过期。
basicmachines-co/basic-memory
Produces PR, changelog and two-week retro images for the Basic Memory repository from evidence in PR bodies, saved to fixed paths under docs/assets/infographics.
krusemediallc/arcads-claude-code
Generate one or more standalone Meta image-ad creatives via ChatGPT Image 2 (gpt-image-2) through the Arcads external API.
Categories
This skill should be used when the user asks to "generate an image", "create a logo", "draw an icon", "edit this photo", "change background to transparent", "remove background", "use GPT image"…. Gpt Image 2 Skill is an agent skill from Wangnov/gpt-image-2-skill. This skill should be used when the user asks to "generate an image", "create a logo", "draw an icon", "edit this photo", "change background to transparent", "remove background", "use GPT image", "use Codex to draw", "用 GPT image 生成图片", "用 Codex 画图", "帮我生成一张图", "改成透明背景", "把这张图编辑一下", or any prompt-to-image or reference-image-edit task that benefits from a structured CLI returning JSON results and JSONL progress events.
Gpt Image 2 Skill fits situations like: asks to generate an image; edit this photo; change background to transparent; remove background.
Run `npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a claude-code`. Or copy the skill folder (skills/gpt-image-2-skill in Wangnov/gpt-image-2-skill) into .claude/skills/gpt-image-2-skill in your project. Claude Code loads it when a task matches its description.
Run `npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a codex`. Or copy the skill folder (skills/gpt-image-2-skill in Wangnov/gpt-image-2-skill) into .agents/skills/gpt-image-2-skill in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Wangnov/gpt-image-2-skill --skill gpt-image-2-skill -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/gpt-image-2-skill, .gemini/skills/gpt-image-2-skill, .github/skills/gpt-image-2-skill and .opencode/skills/gpt-image-2-skill in your project.
Going by SKILL.md and its folder, Gpt Image 2 Skill needs JavaScript for the scripts in its folder, the command-line tools its instructions call (node, npm and cargo) and credentials named OPENAI_API_KEY. Our summary lists: Node.js; A credential in OPENAI_API_KEY.
SKILL.md names 1 domain. As links in the text: developers.openai.com. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.
Gpt Image 2 Skill is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 4.5k tokens (SKILL.md is roughly 18k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 13k tokens, read only when the agent opens those files.
Skills that share tags, products or a category with Gpt Image 2 Skill: Nano Banana Pro Prompts Recommend Skill (YouMind-OpenLab/nano-banana-pro-prompts-recommend-skill, 1.9k stars), Yingzao (op7418/guizang-yingzao-skill, 496 stars), AI Image Creator (centminmod/my-claude-code-setup, 2.7k stars) and Character Refs (eternityspring/shuohao-skills, 4.3k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Wangnov (a GitHub user) maintains it in Wangnov/gpt-image-2-skill, which has 142 GitHub stars. The repository was last updated on October 5, 2026.
Source: Wangnov/gpt-image-2-skill on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.