# Nano Banana

> A Claude skill from Orkas-AI/Orkas-Awesome-AgentSkills.

- **Type:** Skill
- **Install:** `agentstack add skill-orkas-ai-orkas-awesome-agentskills-nano-banana`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [Orkas-AI](https://agentstack.voostack.com/s/orkas-ai)
- **Installs:** 0
- **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [Orkas-AI](https://github.com/Orkas-AI)
- **Source:** https://github.com/Orkas-AI/Orkas-Awesome-AgentSkills/tree/main/creation/skills/nano-banana
- **Website:** https://orkas.ai

## Install

```sh
agentstack add skill-orkas-ai-orkas-awesome-agentskills-nano-banana
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Nano Banana

## When to Use

Use this skill when the user explicitly wants to use Nano Banana, Gemini image models, Imagen-style generation/editing, Google AI Studio, or Atlas Cloud for image execution.

Use for:

- Text-to-image through a configured provider.
- Image editing with one or more local reference/source images.
- Comparing provider availability, privacy, and cost before execution.
- Running a generation/editing task after an image design brief is ready.

Do not use this skill as a generic prompt-writing guide. Use `image-design` for planning and route here only when execution through this provider is appropriate and available.

## Provider And Privacy Rules

1. Check which credentials are configured before choosing a provider.
2. If multiple providers are configured, ask the user which to use unless cost/privacy preference is already clear.
3. If no credentials are configured, provide setup guidance but do not ask the user to paste secrets into chat.
4. Before uploading any local image, ask for explicit confirmation and state the destination provider.
5. State that prompts and uploaded images may be sent to third-party services.
6. Ask for budget limits before batch generation or high-resolution runs.
7. Save outputs to a user-visible local directory and return output paths.

## First Checks

Check environment variables without printing secret values:

```bash
test -n "$ATLASCLOUD_API_KEY" && echo "ATLASCLOUD_API_KEY configured"
test -n "$GEMINI_API_KEY" && echo "GEMINI_API_KEY configured"
```

If the source skill's helper script is available in the local environment, prefer using it. Otherwise use the provider API documented for the user's configured provider.

## Workflow

1. Confirm task type: text-to-image or image edit.
2. Confirm prompt, source/reference image paths, aspect ratio, resolution, output directory, and count.
3. Choose provider:
   - Atlas Cloud when configured and acceptable to the user.
   - Google AI Studio / Gemini when configured and acceptable to the user.
   - no execution when credentials are missing; return setup guidance and prompt/brief instead.
4. Confirm cost, upload, and privacy before execution.
5. Submit the request.
6. Poll or wait for the result if asynchronous.
7. Download or save generated outputs.
8. Verify file existence and return paths plus provider/task metadata.

## Return Format

```markdown
## Nano Banana Result

- Provider:
- Task type:
- Output path(s):
- Task id / remote URL(s), if any:
- Cost / quota note:
- Uploads confirmed:
- Checks performed:
- Failures or fallback:
```

## Limits and Known Issues

- Provider names, model IDs, prices, and availability can change. For production work, verify current provider docs.
- Exact text, logos, faces, product geometry, and reference consistency are not guaranteed.
- Reference image limits depend on the provider and current model.
- This skill must not expose API keys or raw provider responses containing secrets.
- If execution is unavailable, return a prompt and execution plan; do not claim an image was generated.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [Orkas-AI](https://github.com/Orkas-AI)
- **Source:** [Orkas-AI/Orkas-Awesome-AgentSkills](https://github.com/Orkas-AI/Orkas-Awesome-AgentSkills)
- **License:** MIT
- **Homepage:** https://orkas.ai

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-orkas-ai-orkas-awesome-agentskills-nano-banana
- Seller: https://agentstack.voostack.com/s/orkas-ai
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.