# Generate and Edit Images with Nano Banana 2

Creating polished images often requires separate tools for generation, editing, and current information. This skill provides one inference.sh workflow for all three tasks.

## Install

```bash
npx skillstore add inferen-sh/nano-banana-2
```

## Metadata

- Status: approved
- Slug: inferen-sh-nano-banana-2
- Skillstore revision: r2
- Version status: missing
- Tree hash: 9905b4aac8fd971da73e78586a298a5cd29b1b72b15dc550c1ccad0343829ef6
- Author: inferen-sh
- GitHub username: inferen-sh
- License: MIT
- Repository: https://github.com/inferen-sh/skills/tree/main/tools/image/nano-banana-2/
- Ref: 4121de961d1b6f2ffca856260e239505c302452c
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network
- Quality score: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/inferen-sh-nano-banana-2
- Manifest: https://skillstore.pages.dev/api/skills/inferen-sh-nano-banana-2/manifest

## Capabilities

- Generates images from natural-language prompts.
- Edits up to 14 JPEG, PNG, or WebP reference images.
- Requests multiple images in one generation job.
- Controls aspect ratio, resolution, and output format.
- Enables optional Google Search grounding for current information.
- Supports CLI execution and Python SDK streaming.

## Use Cases

- Create Campaign Visuals: Generate several branded concepts with controlled dimensions for social posts, advertisements, or presentations.
- Revise Design References: Edit supplied images while preserving important subjects, composition, and visual details.
- Prototype Image Workflows: Test image generation through the CLI or stream progress through the Python SDK.

## Prompt Templates

### Generate a Basic Image

```
Create an image of [subject] in [style], using [lighting] and a [composition] composition.
```

### Create Multiple Concepts

```
Generate [number] distinct concepts for [project]. Use [aspect ratio], [resolution], and these visual requirements: [details].
```

### Edit Reference Images

```
Using the supplied images, change [element] to [new state]. Preserve [features] and match the original lighting and perspective.
```

### Ground a Current Visual

```
Create a factual visualization of [current topic] for [location or date]. Use Google Search grounding and clearly represent [required facts].
```

## Limitations

- Requires the inference.sh belt CLI or Python SDK and a valid service login.
- Prompts and reference images are processed by an external service.
- Output quality and prompt adherence can vary between requests.
- Google Search grounding must be enabled explicitly and may increase processing requirements.

## Best Practices

- Describe the subject, composition, lighting, style, and required details in one focused prompt.
- Use only necessary reference images and remove sensitive information before upload.
- Generate lower-resolution concepts before requesting final 4K outputs.

## Anti Patterns

- Do not use vague prompts when layout, colors, or subject details are important.
- Do not upload confidential images or personal data to external processing services.
- Do not enable Google Search grounding when the request does not require current information.

## Security Audit

- Audited at: 2026-08-06T10:43:30.748\+00:00
- Summary: All 38 static findings are false positives caused by Markdown formatting, fixed CLI examples, placeholder URLs, or documentation links. The skill intentionally uses the belt CLI for remote image inference, but no hidden execution, prompt injection, or credential collection was found.

## Stats

- Views: 206
- Downloads: 12
- Favorites: 0
- Popularity score: 0