# Generate AI Images with inference.sh

Creating visual assets across many image models is slow when each provider has different steps. This skill gives Claude, Codex, and Claude Code practical inference.sh CLI examples for generation, editing, upscaling, and stitching.

## Install

```bash
npx skillstore add inference-sh/ai-image-generation
```

## Metadata

- Status: approved
- Slug: inference-sh-ai-image-generation
- Skillstore revision: r1
- Version status: missing
- Tree hash: d1a46d88de7eebc861f177ed540fd2cb5938788ceb87c5b9208070d521c597b8
- Author: inference-sh
- GitHub username: inference-sh
- License: MIT
- Repository: https://github.com/inference-sh/skills/tree/main/skills/ai-image-generation/
- Ref: a06681402992ceae98ba04d54cfd4ab004862696
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: blocked
- Manual install advisory: allowed\_with\_warning
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/inference-sh-ai-image-generation
- Manifest: https://skillstore.pages.dev/api/skills/inference-sh-ai-image-generation/manifest

## Capabilities

- Shows inference.sh commands for text-to-image generation with FLUX, Gemini, Grok, Seedream, and Reve models.
- Lists image model identifiers and their intended strengths for faster model selection.
- Provides examples for product photos, landscapes, posters with text, portraits, and cyberpunk scenes.
- Includes image upscaling and multi-image stitching examples through inference.sh apps.
- Supports CLI workflows in Claude, Codex, and Claude Code when inference.sh access is available.

## Use Cases

- Create Concept Art: Generate visual directions for characters, scenes, posters, and product ideas using model-specific examples.
- Produce Marketing Visuals: Draft social graphics, campaign images, and product mockups with repeatable prompt structures.
- Prototype Image Workflows: Test inference.sh image apps from a terminal before adding them to scripts or product experiments.

## Prompt Templates

### Generate a Simple Image

```
Generate an image of a calm mountain lake at sunrise. Recommend one inference.sh model and explain the expected style.
```

### Choose a Model for a Task

```
I need a product mockup with clean studio lighting and readable label text. Compare two suitable models and recommend one.
```

### Upscale or Edit an Image

```
Use an existing approved image URL to plan an upscale or edit. Include privacy cautions and the model choice.
```

### Plan a Repeatable Asset Workflow

```
Create a repeatable workflow for generating five branded social images with consistent style, aspect ratios, and review checkpoints.
```

## Limitations

- Requires the inference.sh CLI, an account login, and network access to remote model providers.
- Model availability, pricing, latency, and output quality are controlled outside the skill.
- Prompts and image URLs may be sent to external services during generation or editing.
- Does not provide local image safety checks, rights review, or asset storage management.

## Best Practices

- Define subject, style, aspect ratio, lighting, and text needs before choosing a model.
- Use only approved image URLs and avoid sending confidential assets to remote services.
- Verify installation sources and account permissions before running inference.sh commands.

## Anti Patterns

- Do not paste private customer images, unreleased designs, or secrets into image prompts or URLs.
- Do not run remote installer scripts without source verification and checksum or signature checks.
- Do not launch many high-cost generations without confirming budget, resolution, and review limits.

## Security Audit

- Audited at: 2026-07-06T15:48:33.91\+00:00
- Summary: AI review confirmed the pipe-to-shell installer on SKILL.md line 22 as a critical risk. The related network URL on line 22 is also risky because it downloads executable installer content. The remaining backtick and URL findings are Markdown documentation, model identifiers, or placeholder examples with no prompt injection or covert exfiltration evidence.

## Stats

- Views: 246
- Downloads: 32
- Favorites: 0
- Popularity score: 0