Hubbestaiskill.com

Home / All Skills / Content & Media Creation / ai-image-generation

Content & Media Creation · 101-skills/superpowers

ai-image-generation

Generate AI images with GPT-Image-2, FLUX, Gemini, Grok, Seedream, Reve and 50+ models via inference.sh CLI. Models: GPT-Image-2, FLUX Dev LoRA, FLUX.2 Klein LoRA, Gemini 3 Pro Image, Grok Imagine, Seedream 4.5, Reve, ImagineArt. Capabilities: text-to-image, image-to-image, inpainting, LoRA, image editing, upscaling, text rendering. Use for: AI art, product mockups, concept art, social media graphics, marketing visuals, illustrations. Triggers: flux, image generation, ai image, text to image, stable diffusion, generate image, ai art, midjourney alternative, dall-e alternative, text2img, t2i, image generator, ai picture, create image with ai, generative ai, ai illustration, grok image, gemini image, gpt image, openai image, chatgpt image

L4 High-Privilege · Graded auth & full auditV1 Commit-PinnedExecute CommandNetwork AccessDual-pass Audit

Author 101-skillsGitHub 101-skills/superpowers ↗License MIT（仓库 README 徽标 'License: MIT' 与 .claude-plugin/plugin.json 中 "license": "MIT" 声明；但 pin 树内无 LICENSE 文件，GitHub API 对两 fork 亦报无 license —— 授权只出现在元数据/徽标，无正式许可文本）Commit becc256497

Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.

1How It Works & Mechanism

Shares the same framework with other tool skills: whitelisted execution via belt CLI + a catalog table of '50+ models' (16 app-id rows in the table: GPT-Image-2 / FLUX series / P-Image series / Gemini / Grok / Seedream / Reve / ImagineArt / Topaz), where an agent triggers remote inference simply by filling the --input JSON.

tools/image/ai-image-generation/SKILL.md

belt app run falai/flux-dev-lora --input '{"prompt": "a cat astronaut in space"}'

Quick Start is completely isomorphic to the twitter/video series—tool skills in this repository are essentially different 'catalog pages' of the same template.

Expresses 'multiple capabilities within the same model' by toggling input keys: passing only prompt to gpt-image-2 performs text-to-image, while appending images:[...] triggers image editing/inpainting; meanwhile, apps like P-Image-Edit and Reve serve as standalone apps for editing.

tools/image/ai-image-generation/SKILL.md

"prompt": "change the background to a beach at sunset",

The input contract has no schema file and relies entirely on example JSON demonstrations—the model infers parameter keys (quality/images/size/aspect_ratio/width/height) from these examples.

Differentiated selling points are authored as reproducible examples: text rendering (Reve's 'HELLO WORLD' poster), 4K tier (Seedream 4.5), and ultra-low cost (FLUX Klein $0.0001/image) each come with a command snippet for agents to select models on demand.

tools/image/ai-image-generation/SKILL.md

"prompt": "A poster that says HELLO WORLD in bold letters"

Selection rationales (ultra-cheap / 2K-4K / text rendering) are directly noted within the table rows, forming a routing decision layer.

Post-processing/utility apps are unified into the same entry point: Topaz upscaling and stitch-images are listed side-by-side with generative models in the table, indicating that 'image-to-image post-processing' has also been app-ified by the platform.

tools/image/ai-image-generation/SKILL.md

belt app run infsh/stitch-images --input '{

Stitching/upscaling requires no local image tools outside of this skill, further minimizing the need for local dependencies.

Positioned within the ecosystem as an 'asset source for content pipelines': in addition to peer image packages (p-image/gpt-image/flux-image/upscaling/background-removal), Related Skills interlink with ai-video-generation (inputs for image-to-video), ai-avatar-video (avatar assets), and twitter-automation (posting images).

tools/image/ai-image-generation/SKILL.md

npx skills add inference-sh/skills@background-removal

The README also lists this skill as one of the representative items in the 'Tools' group, alongside video and search.

2Core Capabilities

01Text-to-image: GPT-Image-2, FLUX Dev/Klein, Gemini 3 Pro/2.5 Flash, Grok, Seedream, Reve, ImagineArt

02Image editing / inpainting (gpt-image-2 via images[], P-Image-Edit, Reve natural language editing)

03LoRA styling (preset LoRA styles via flux-dev-lora/flux-2-klein-lora/p-image-lora)

04Model selection for key selling points like 4K/text rendering (Seedream 4.5 2K-4K, Reve text rendering)

05Post-processing: Topaz image upscaling, stitch-images multi-image stitching

06Asset generation chained with video, avatar, and social media publishing skills

3External Dependencies

Kind	Dependency / Endpoint
cli	belt（inference.sh CLI）
package	belt-sh/cli（npx skills add 安装）
api	平台图像模型/工具 app：openai/gpt-image-2、falai/flux-dev-lora、falai/flux-2-klein-lora、pruna/p-image(-lora/-edit)、google/gemini-3-pro-image-preview、xai/grok-imagine-image、bytedance/seedream-4-5、falai/reve、falai/imagine-art-1-5-pro-preview、pruna/flux-klein-4b、falai/topaz-image-upscaler、infsh/stitch-images
network	平台文档（docs/apps/running、image-generation example）
network	安装说明托管

4Recommended Governance & Evidence

Suggested Controls for Safe Operation

For Agent Runtime

Graded/per-use authorization, sandbox recommended

For Administrators

Approval workflow, rollback support, full audit

Evidence & Triggers (File Paths & Excerpts)

网络端点: Declares external network access: platform belt CLI → inference.sh platform (image generation/editing apps; outputs hosted on cloud.inference.sh) [
SKILL.md: Declares invocation of host CLI interpreters or external execution environments

Audit Notes: . Consistent with other tool skills in this repository: contains no scripts, no credential-reading logic, and no direct networking code; execution is routed strictly via the whitelisted belt CLI with predictable outbound destinations (inference.sh and its underlying image model providers). Points to note: ① Every generation consumes account credits, especially higher-tier models (4K / Pro preview models); ② Reference and output images leave the local network to third parties, offering no confidentiality guarantees for assets; ③ Supply chain risks mirror preceding skills (unpinned npx, unrenamed inference-sh mirror, missing LICENSE file). Per unified classification guidelines: the `belt` invoked by this skill is a third-party npm/npx-distributed CLI (installed via `npx skills add belt-sh/cli`), runs under a logged-in state via `belt login`, and consumes inference.sh account quota/credits—falling under 'invokes third-party CLI and touches quota/login state', and is therefore classified as Orange (use after evaluation); the skill source code itself contains no token or secret retrieval logic.

5Second-Pass Verification

[discrepancy] description 宣称 50+ models — 表内可数 app-id 行为 16，'50+ models' 在本 SKILL.md 无法定位核实，属平台注册表层面宣称（belt app list --category image 可查全量）。
[ok] Models 名单逐项定位 — GPT-Image-2/FLUX Dev LoRA/FLUX.2 Klein/Gemini 3 Pro/Gemini 2.5 Flash/Grok Imagine/Seedream 4.5/Reve/ImagineArt 均在表中有行；description 提到 'Seedream 4.5' 而表内另含 4.0/3.0，无冲突。
[ok] 能力声明示例支撑（i2i/inpainting/LoRA/editing/text rendering/upscaling） — 编辑由 gpt-image-2 的 images[] 示例与 P-Image-Edit/Reve 支撑；LoRA 由 flux-dev-lora/flux-2-klein-lora/p-image-lora 支撑；text rendering 由 Reve HELLO WORLD 示例支撑；inpainting 仅在 description/表行出现，无独立示例（同 gpt-image-2 编辑能力，未单列）——记为部分 ok。
[ok] 无脚本/无凭证/无直连网络 — 目录仅 SKILL.md；curl/wget/密钥 token 零命中；执行面 allowed-tools: Bash(belt *)。
[ok] 本地文件写入面 — 本文件示例无 shell 重定向，不写本地文件。
[ok] 图像输入全远程 URL — images/image_url 值均为 https://…；无本地上传/路径读取指令。
[ok] license/stars/commit 元数据 — pin == 本地 HEAD == becc25649700；101-skills/superpowers stars=0、pushed_at 2026-08-31T22:58:53Z；MIT 声明见 plugin.json/README 徽标，树内无 LICENSE 文件。
[unlocatable] installs 安装量 — 无公开安装量数据源，置 null。

6Verdict & Scenarios

A single entry point covers mainstream model families: image models from OpenAI, fal, Google, xAI, and ByteDance are unified into a single belt invocation surface, eliminating the need to integrate SDKs individually.

Front-loaded cost anchoring: low-cost tiers like FLUX Klein at $0.0001/image are documented in the table, enabling cost-sensitive users to select their tier.

Reproducible selling points: text rendering, 4K, editing, and LoRA each provide minimal examples, allowing agents to copy them directly for reliable results.

Ecosystem interlinking enables an end-to-end flow from 'image generation → video generation → post publishing'; this skill serves as the asset source node in content pipelines.

Best for: Ideal for marketing, design, and content teams to batch-generate and edit images (product shots, social media graphics, concept art) within agent workflows; suitable for users who need cross-model comparisons or want specialized capabilities like LoRA, 4K, or text rendering; also serves as an asset source for downstream skills such as landing pages, video, and avatars.

Not for: Not suitable for scenarios requiring images to stay within an intranet or be generated strictly locally; not suitable for high-frequency free trials (all usage is pay-as-you-go); overkill for lightweight users without multi-model switching needs, who are better off using standalone official products (e.g., ChatGPT, Gemini); not suitable for enterprise compliance scenarios requiring fine-grained control over prompt safety and content moderation policies.

Install Agent-Ready

① Verified Mirror 2026-09-09

Option A · Direct DownloadDownload ai-image-generation.tar.gz

sha256: f66b680166c1a445…

Option B · Copy Guide for Agent

Agents automatically download and verify mirrors from this JSON guide.

② Upstream GitHub

GitHub Repo ↗

Pinned to commit becc256497.