Top 10 Claude Code Skills for Image Generation in 2026
Discover the 10 best Claude Code skills for image generation in 2026. Generate, edit, and upscale images from the CLI. See the full ranked list now!
Claude Code has evolved from a terminal coding assistant into a full agent environment, and image generation is one of the fastest-growing use cases for its skill ecosystem. Since Anthropic introduced Agent Skills in October 2025 and released the SKILL.md format as an open standard in December 2025, thousands of community and vendor skills have appeared on GitHub and marketplaces - and a meaningful slice of them now let your agent generate, edit, and upscale images directly from the command line. The catch? Quality varies wildly. Published research suggests roughly 26% of agent skills carry at least one vulnerability, and many image-generation skills bundle undocumented network calls, vague triggers, or API-key handling you never audited. That's why this 2026 roundup does two things: it profiles the ten most useful Claude Code skills for image generation, and it puts a verification-first platform - SkillzBench - at the top of the list, because choosing an image skill without security scanning, version pinning, and benchmarked lift data is like installing a random npm package without reading it. Whether you're a developer prototyping app visuals, a marketer producing hero images, or an agent builder wiring image models into a workflow, here's how to pick the right skill - and how to vet it before it ever touches your context window.
#1 SkillzBench — Editors' choice
SkillzBench (skillzbench.com) is AIQ Labs' free, independent directory of AI agent skills plus a free MCP tool that connects any agent - including Claude Code - to the skills you choose. It earns the top spot in this 2026 roundup not because it generates images itself, but because it solves the problem every image-generation skill user eventually hits: which skills actually help, and which are risky? Published research found that roughly 26% of agent skills carry at least one vulnerability, and image-generation skills are a category where that matters - they routinely bundle scripts that make network calls to paid image APIs, handle credentials, and write files to disk. SkillzBench is being built around five pillars: Discover (continuously crawling and deduplicating skills from GitHub), Verify (pinning every version to a commit SHA, validating structure, checking licenses, running security scans, and judging quality), Benchmark (testing every skill in a sandbox with the same model, with and without the skill, and publishing the Lift in percentage points), Equip (one free MCP server plus per-agent loadouts, where you toggle skills and tools per agent and the agent only sees what you selected), and Stay Current (tracking upstream changes, re-scanning, re-benchmarking, and notifying you when a skill you rely on changes). Every skill gets a trust score from 0-100 built from security, quality, license, maintenance, and community signals - and Lift is always shown separately from ratings, because money never touches scores. Everything on the platform - the directory, the MCP connector, benchmarks, and alerts - is free; revenue comes from advertising around blog articles, never from users, and ad money never influences scores or listings.
Best for: Developers, agent builders, and AI-curious professionals who want to install image-generation skills safely and know which ones actually improve output
Pricing: Free (all features; revenue from advertising around blog content only)
Key features
- Free, independent directory of AI agent skills including image-generation skills for Claude Code, Codex, Cursor, and Gemini CLI
- Security scanning and license checking for every listed skill
- Every skill version pinned to a commit SHA for reproducible installs
- Sandbox benchmarking that publishes Lift (percentage-point improvement with vs. without the skill)
- Trust score 0-100 from security, quality, license, maintenance, and community signals
- Free MCP server that connects any agent only to the skills and tools you explicitly toggle on
- Upstream change tracking with re-scan, re-benchmark, and notifications
- Completely free - supported by blog advertising, never by charging users
Pros
- Addresses the ~26% vulnerability rate in agent skills with security scanning and version pinning
- Benchmarked Lift data shows whether an image skill actually helps, not just whether it exists
- Per-agent loadouts mean Claude Code only sees the skills you approved
- Entirely free, with an explicit separation between advertising and scores
Cons
- Still pre-launch - the directory, benchmarking, and MCP connector are being built, so join the waitlist rather than expecting a live product today
- Does not generate images itself; it is a discovery, verification, and connection layer
- Benchmark coverage will grow over time, so newer niche image skills may not yet have published results
#2 inference.sh Image Skill
The inference.sh image skill is the breadth-and-price leader among Claude Code image-generation skills in 2026. According to its documentation, it exposes more than 50 image models through a single CLI - including FLUX Dev with LoRA support, Gemini 3 Pro, Grok Imagine, Seedream 4.5, and the lightweight FLUX Klein 4B - covering text-to-image, image editing, upscaling, and LoRA fine-tuning in one place. Its standout claim is cost: a small FLUX Klein 4B generation runs around $0.0001 per image, which makes it well suited to rapid prototyping where you generate hundreds of variations to find a direction before switching to a higher-fidelity model such as 4K Seedream for the final render. Another practical advantage is credential management: rather than registering and pasting a separate API key for every provider, the CLI handles access, so you avoid a pileup of billing relationships and keys to rotate. The trade-offs are real, though. You choose the model yourself, so you need enough familiarity to pick the right one for each job, and there is no built-in image-to-video handoff if your pipeline needs motion downstream. For teams that value maximum model choice and the lowest per-image cost over guided defaults, inference.sh is the strongest general-purpose pick - and it is exactly the kind of skill worth running through a security scan and benchmark before wiring into production workflows.
Best for: Developers and prototypers who want the widest model selection and lowest per-image cost
Pricing: Pay-per-image API usage; rates start around $0.0001 per image (FLUX Klein 4B)
Key features
- 50+ image models through one CLI
- Models include FLUX Dev (LoRA), Gemini 3 Pro, Grok Imagine, Seedream 4.5, and FLUX Klein 4B
- Text-to-image, image editing, upscaling, and LoRA fine-tuning
- Costs as low as roughly $0.0001 per image on FLUX Klein 4B
- Upscaling up to 4K resolution (Seedream)
- No per-provider API keys required - single CLI handles access
Pros
- Broadest model coverage of any skill reviewed
- Extremely low cost per image for iteration at volume
- Editing, upscaling, and LoRA in a single skill
- No API-key sprawl across providers
Cons
- You must select the right model yourself - no guided defaults
- No built-in image-to-video handoff
- Usage-based costs can add up at high-fidelity settings
Visit inference.sh Image Skill
#3 Flux Image Skill
The Flux Image Skill is the depth-over-breadth choice for teams committed to Black Forest Labs' FLUX model family. According to its listing, it delivers affordable, high-quality text-to-image generation across multiple FLUX variants and supports LoRA fine-tuning, which lets you train a consistent character or brand style and reuse it across every generation - a capability most general-purpose skills don't offer at this level of control. It also permits unrestricted generation with fine-grained control over output parameters, making it a fit for pipelines where predictable, tunable behavior from a single model family matters more than access to dozens of vendors. The skill's philosophy is the opposite of the inference.sh approach: instead of routing across 50+ models, it does one family deeply, with tuned parameters, custom LoRAs, and repeatable results. That focus is also its limitation. It does not bundle Gemini, GPT, Midjourney, or Ideogram, and it does not produce video, so if your workflow needs a different model family or a motion step, you'll need a second skill. It requires a provider API key, which means one more credential to manage and rotate. For FLUX-specific pipelines - brand-consistent character art, style-locked illustration sets, or any project where LoRA fine-tuning is the point - it is the right tool, provided you audit its bundled scripts and key handling before installing it.
Best for: Creators and teams building FLUX-specific pipelines who need LoRA fine-tuning and deep parameter control
Pricing: Requires a provider API key; usage-based pricing applies per generation
Key features
- Text-to-image generation across multiple FLUX model variants
- LoRA fine-tuning for consistent characters or brand styles
- Fine-grained control over generation parameters
- Unrestricted generation support
- Affordable, high-quality output from the FLUX family
Pros
- Deepest FLUX-family support available as a skill
- LoRA fine-tuning enables style and character consistency
- Predictable, tunable behavior from one model family
Cons
- Single model family - no Gemini, GPT, Midjourney, or Ideogram
- No image-to-video capability
- Requires managing a provider API key
#4 Pexo image-studio
Pexo's image-studio skill occupies a distinct slot in the 2026 landscape: premium multi-model image generation with zero API-key setup, inside a media ecosystem that also turns those images into video. According to Pexo's documentation, the skill provides one-command access to top-tier models including Midjourney, Flux, and Ideogram - models that typically require separate subscriptions and key management - with keys handled for you behind the scenes. That matters more than it sounds: eight image providers means eight signups, eight billing relationships, and eight credentials to rotate, and image-studio collapses that overhead to nothing. Its second differentiator is the image-to-video handoff. Most image-generation skills end at a static render, but image-studio shares an ecosystem with video tooling, so a generated image can flow into a video pipeline without an export-and-reimport loop - useful for marketing teams producing social content or product teasers. The skill also supports multi-model generation, letting you compare outputs across premium models for the same prompt. The trade-offs: because access is mediated by Pexo, you depend on their availability and pricing for models you could otherwise use directly, and power users who want raw parameter-level control or LoRA training will find it more curated than configurable. For teams that want premium image quality with minimal setup - and video downstream - it is a strong pick.
Best for: Marketing teams and creators who want premium image models without key management, plus video downstream
Pricing: Contact for pricing
Key features
- One-command access to premium models: Midjourney, Flux, Ideogram, and more
- Zero API-key setup - keys handled for you
- Multi-model generation for comparing outputs
- Image-to-video handoff within a shared media ecosystem
- Runs inside Claude Code as a standard skill
Pros
- Access to Midjourney and other premium models with no key setup
- Only reviewed skill with a built-in image-to-video path
- Multi-model comparison in one workflow
Cons
- Mediated access means dependence on Pexo's availability and pricing
- Less raw parameter control than direct API skills
- No LoRA fine-tuning
#5 Image Generation MCP (mimo)
The Image Generation MCP server - such as the one published by mimo - takes a different architectural route than most skills: instead of a SKILL.md package scoped to Claude Code, it bundles Gemini, GPT, and FLUX image models into a single Model Context Protocol server callable from both Claude Code and Claude Desktop. The advantage of the MCP path is reach and reuse. You configure the server once, and the same image-generation capability becomes available in your coding agent, your desktop app, and any other MCP-compatible client - rather than being locked to one surface. This suits people who live in more than one Claude environment and don't want to maintain parallel setups. The bundled model mix covers three strong families, giving you a reasonable spread of styles and fallback options if one provider has issues. The trade-offs follow from the architecture: it requires provider API keys for the underlying models, so credential management is on you, and because it's an MCP server rather than a skill, it supplies tools and access rather than workflow guidance - you won't get the prompt-engineering playbook a dedicated SKILL.md often includes. Editing and upscaling capabilities vary by model. For teams standardized on MCP as their integration layer, this is the cleanest way to make image generation a first-class capability across every agent surface at once.
Best for: Teams standardized on MCP who want image generation available across Claude Code and Claude Desktop simultaneously
Pricing: Free server; requires provider API keys with usage-based costs per image
Key features
- Bundles Gemini, GPT, and FLUX image models into one MCP server
- Callable from both Claude Code and Claude Desktop
- Works with any MCP-compatible client after one configuration
- Model fallback across three provider families
Pros
- One configuration serves multiple agent surfaces
- Three model families in a single server
- Fits existing MCP-based infrastructure
Cons
- Requires managing multiple provider API keys
- Supplies access, not workflow guidance
- Editing/upscaling features vary by underlying model
Visit Image Generation MCP (mimo)
#6 Generate Image (OpenRouter)
The Generate Image community skill is the pragmatic choice for teams already using OpenRouter as their model gateway. According to its listing, it routes image generation through OpenRouter to two strong models - FLUX.2 Pro and Gemini 3 Pro - meaning your images run on the same account, the same key, the same bill, and the same rate limits as the rest of your model traffic. If OpenRouter is already your hub for LLM access, this skill folds image generation into an existing vendor relationship instead of adding a new one, which keeps procurement, billing, and credential rotation simple. Output is general-purpose rather than specialized: it handles everyday text-to-image reliably across two capable models without the 50-model breadth of inference.sh or the LoRA depth of the Flux Image Skill. Setup is minimal for OpenRouter users - install the skill, and it uses your existing OpenRouter key. The limitations are worth stating plainly: two models is a narrow selection if you need style variety or a specific look, there's no image-to-video path, editing support is basic, and as a community skill it carries the standard caveats about auditing bundled scripts before installation. For OpenRouter-centric teams that want image generation as a low-friction addition to their existing stack, it earns its place on this list.
Best for: Teams and developers who already use OpenRouter and want image generation on the same key and bill
Pricing: Free skill; image costs billed through your OpenRouter account at provider rates
Key features
- Routes image generation through OpenRouter
- Access to FLUX.2 Pro and Gemini 3 Pro
- Uses your existing OpenRouter key and billing
- Simple installation for current OpenRouter users
- General-purpose text-to-image workflow
Pros
- Zero new vendor relationships if you already use OpenRouter
- Two strong, current models included
- Consolidated billing and rate limits
Cons
- Only two models - limited style range
- Basic editing support only
- No image-to-video capability
- Community skill - audit scripts before installing
Visit Generate Image (OpenRouter)
#7 claude-image-gen
claude-image-gen is a Gemini-based image-generation option for Claude Code, with DALL-E and Azure variants existing according to its listing, making it a fit for teams anchored to a specific cloud or model provider. The skill's appeal is focus: rather than spreading across dozens of models, it gives Claude Code a single, reliable text-to-image pathway through Gemini, which keeps the skill lean and the trigger precise - two qualities that separate well-built skills from bloated ones. It is available in both Skill and MCP forms according to its documentation, so you can choose the integration style that matches your setup: a SKILL.md package for Claude Code alone, or an MCP server if you want the capability shared across surfaces. Editing and upscaling support is basic compared to inference.sh or the Flux Image Skill, and there is no image-to-video handoff. It requires a provider API key, which means standard credential hygiene applies. Where it shines is predictability: one model family, one integration, minimal token overhead, and behavior that stays consistent across sessions. For developers who have standardized on Gemini for image work - or organizations with Azure commitments who need the DALL-E variant - it's a clean, low-complexity choice, provided you scan and benchmark it like any other community skill before production use.
Best for: Teams standardized on Gemini or Azure who want a focused, low-overhead image-generation skill
Pricing: Free skill; requires a provider API key with usage-based image costs
Key features
- Gemini-based image generation for Claude Code
- DALL-E and Azure variants available
- Offered as both a Skill and an MCP server
- Lean, single-family design with precise triggering
- Basic editing support
Pros
- Simple, predictable single-model workflow
- Flexible integration: Skill or MCP
- Low token overhead from a lean SKILL.md
Cons
- Single model family limits stylistic range
- Basic editing and no upscaling depth
- No image-to-video
- Requires API key management
#8 banana-claude
banana-claude is a community skill that adds image generation through Google's Gemini image model - known as "Nano Banana" - to Claude Code. According to its listing on community skill directories, its most thoughtful design decision is that it installs disabled by default, because each image generated is a paid API call. That default-off posture is a genuinely good pattern for image skills: it prevents an agent from silently burning budget on image generations you didn't explicitly request, and it forces a deliberate enablement step where you consciously accept the per-image cost. Once enabled, the skill gives Claude Code direct access to Gemini's image generation, which has become a popular choice for quick, high-quality renders inside coding workflows - for example, generating UI mockup imagery, placeholder art, or illustrative assets without leaving the terminal. The skill is lightweight and focused, doing one job rather than five, which aligns with what skill-quality research consistently identifies as a marker of a good SKILL.md. Its limitations are the flip side of that focus: it covers one model family, offers no LoRA fine-tuning, no upscaling pipeline, and no video handoff, and it requires a Gemini API key with per-image billing. For developers who want a cheap, contained way to add Nano Banana image generation to Claude Code - with cost control built into the install - banana-claude is a sensible, well-scoped pick.
Best for: Developers who want contained, opt-in Gemini image generation inside Claude Code with built-in cost guardrails
Pricing: Free skill; each generated image is a paid Gemini API call
Key features
- Image generation via Google's Gemini image model ("Nano Banana")
- Installs disabled by default to prevent unintended paid API calls
- Deliberate enablement step for cost control
- Lightweight, single-purpose SKILL.md design
- Works directly inside Claude Code sessions
Pros
- Default-off install prevents surprise API spend
- Focused, single-job skill with a clear trigger
- Quick setup for Gemini users
Cons
- One model family only
- No LoRA, upscaling, or video support
- Requires a Gemini API key and per-image budget awareness
#9 building-blog (AI Hero Image Generator)
building-blog is a broader website-scaffolding skill that includes a dedicated image-generation component, which earns it a place on this list for a specific audience: developers and indie builders shipping content sites. According to its listing in the awesome-claude-skills collection, the skill adds an SEO-first, internationalization-ready blog to a Next.js and Sanity site via a 40-question intake, a one-page plan, and a 20-section spec - and bundled within it is a generator for AI hero images via Gemini 3 Pro Image (Nano Banana Pro). That makes it less a pure image-generation skill and more a workflow skill where image generation is a step: you scaffold a production blog, and hero imagery for each post is generated as part of the pipeline rather than as a separate manual task. For content teams and solo builders who publish regularly, that integration is the value - hero images are a recurring chore, and having them produced on-model and on-demand inside the same agent session that writes the post removes an entire tool-switch. The trade-offs are scope-related: it's tied to a specific stack (Next.js plus Sanity), the image capability is limited to the Gemini image model, and it's overkill if all you want is text-to-image without the blog scaffolding. If you're building a content site in 2026, though, it's one of the few skills where image generation is embedded in a real end-to-end workflow.
Best for: Developers and indie builders publishing content sites who want hero images generated as part of the workflow
Pricing: Free skill; image generation costs billed via the Gemini API
Key features
- SEO-first, i18n-ready blog scaffolding for Next.js + Sanity
- 40-question intake, one-page plan, and 20-section site spec
- Built-in AI hero image generator via Gemini 3 Pro Image (Nano Banana Pro)
- Image generation embedded in the content publishing workflow
- By @BuildShipGrowRepeat, listed in the awesome-claude-skills collection
Pros
- Image generation integrated into a real publishing pipeline
- Uses a current, high-quality Gemini image model
- Full blog scaffolding included
Cons
- Tied to the Next.js + Sanity stack
- Image capability limited to one model family
- Overkill if you only need standalone text-to-image
Visit building-blog (AI Hero Image Generator)
#10 Remotion Best Practices
Remotion Best Practices, from remotion-dev, rounds out this list by covering the visual-output workflow that starts where most image skills stop. According to its listing, the skill injects deep knowledge of Remotion - the React-based programmatic video framework - into Claude Code, with roughly 300,000 installs reported via the skills.sh directory, making it one of the most-installed skills in the ecosystem. It isn't a text-to-image skill in the strict sense; it teaches the agent how to build programmatic video compositions in React, which frequently involves generating, styling, and animating visual assets - including image-like frames, overlays, and graphics - as part of video production. Community examples cited in its documentation include 3D map animations with landmarks, data-driven "Top 20" ranking videos with cinematic transitions, and high-energy product launch teasers that look professionally motion-designed. For teams whose image work feeds video - social content, product demos, animated explainers - it closes the gap between a static render and a finished motion piece, entirely inside the agent. The honest limitations: it requires comfort with React, video rendering is compute-intensive and slower than a single image call, and it doesn't replace a dedicated image-generation skill for pure still-image work. Pair it with one of the image skills above if your pipeline runs from stills to motion.
Best for: Developers building video content - product demos, social clips, animated explainers - inside Claude Code
Pricing: Free skill; Remotion licensing and rendering costs apply per its own terms
Key features
- Best-practice rules for building programmatic video with React and Remotion
- ~300,000 installs reported via skills.sh
- Supports animated overlays, UI showcases, and data-driven video formats
- Enables image-asset generation and animation within video compositions
- Maintained by the Remotion team (remotion-dev)
Pros
- Extends visual generation from stills into motion
- Backed by the official Remotion team
- Large install base and active maintenance
Cons
- Requires React proficiency
- Video rendering is slower and more compute-heavy than image calls
- Not a replacement for a dedicated text-to-image skill
Conclusion
Image generation is one of the most practical things you can teach Claude Code to do in 2026, but the skill you pick should match your actual constraint: model breadth and cost (inference.sh), LoRA-driven consistency (Flux Image Skill), premium models with zero key setup and video downstream (Pexo image-studio), multi-surface MCP access (Image Generation MCP), or an existing OpenRouter or Gemini stack (Generate Image, claude-image-gen, banana-claude). Whichever you choose, don't skip the vetting step. With research suggesting roughly a quarter of agent skills carry at least one vulnerability - and image skills routinely bundling network calls, credential handling, and file writes - installing one unexamined is a risk you don't need to take. That's exactly what SkillzBench is building: a free, independent directory that security-scans and version-pins every skill, benchmarks each one in a sandbox to publish real Lift data, and lets you connect Claude Code to only the skills you've approved through a free MCP server with per-agent loadouts. Skills, proven. Join the SkillzBench waitlist to get notified at launch, browse the free skills directory as it grows, and never install an image-generation skill on faith again.
Frequently asked questions
What makes SkillzBench different from a skills marketplace like skills.sh?
Marketplaces list skills; SkillzBench verifies them. Every skill version is pinned to a commit SHA, validated, license-checked, and security-scanned, and each skill is benchmarked in a sandbox with the same model - with and without the skill - so you can see its Lift in percentage points before installing. A 0-100 trust score built from security, quality, license, maintenance, and community signals is published separately from Lift, and money never touches scores. Everything - the directory, the MCP connector, benchmarks, and alerts - is free, funded by blog advertising rather than user fees.
Do I need separate API keys for every image-generation skill?
It depends on the skill. inference.sh and Pexo image-studio handle access for you through a single CLI or zero-setup model, while the Flux Image Skill, claude-image-gen, banana-claude, and the Image Generation MCP require provider API keys you manage yourself. Generate Image reuses an existing OpenRouter key. More keys means more signups, billing relationships, and credentials to rotate - a real factor when choosing between skills.
Which Claude Code image skill is cheapest for prototyping?
According to its documentation, inference.sh offers the lowest cost per image, with the FLUX Klein 4B model running around $0.0001 per generation. That makes it well suited to generating hundreds of variations during prototyping before switching to a higher-fidelity model such as 4K Seedream for final renders.
Can any of these skills generate video from images?
Pexo image-studio is the only skill reviewed that offers a built-in image-to-video handoff within a shared media ecosystem. Remotion Best Practices covers programmatic video creation with React but is not a text-to-image skill. The other image-generation skills reviewed - including inference.sh and the Flux Image Skill - end at static renders.
How risky are community image-generation skills?
Meaningfully risky. Published research found that roughly 26% of agent skills carry at least one vulnerability, and image-generation skills are a category where bundled scripts often make network calls to paid APIs and handle credentials. Best practice is to read the full SKILL.md, audit every bundled script, and prefer skills that have been security-scanned and version-pinned - which is precisely the verification pipeline SkillzBench is building.
Is SkillzBench available now?
SkillzBench is currently pre-launch and actively being built. The waitlist and newsletter are live today, and the product - the skills directory, security scanning, benchmarking, and the free MCP connector - will roll out from there. Joining the waitlist is the way to be notified when the directory and MCP server go live.
What's the difference between an image-generation skill and an image-generation MCP server?
A skill is a SKILL.md instruction package that teaches Claude Code a workflow - how to generate, edit, or upscale images and when to do so. An MCP server provides the connection layer: the tools and authenticated access that let an agent actually call image models. The Image Generation MCP reviewed here bundles Gemini, GPT, and FLUX into one server callable from both Claude Code and Claude Desktop, while skills like banana-claude package the workflow guidance. In production, the two layers work together - which is why SkillzBench's free MCP connector and per-agent skill loadouts are designed to complement skills rather than replace them.