The skill executes arbitrary remote code from external sources
Claims to do
nano-banana: AI image generation CLI. Default model: Gemini 3.1 Flash Image Preview (Nano Banana 2).
Actually does
The skill installs a CLI tool named `nano-banana` by cloning `https://github.com/kingbootoshi/nano-banana-2-skill.git` and using `bun` to install dependencies and link the executable. It configures a Gemini API key in `~/.nano-banana/.env`. The `nano-banana` command then generates images via the Gemini API, optionally processing them with `ffmpeg` and `imagemagick` for transparency, and logs costs to `~/.nano-banana/costs.json`.
/plugin marketplace add kingbootoshi/nano-banana-2-skill/plugin install nano-banana@kingbootoshi/nano-banana-2-skillnpx skills add https://github.com/kingbootoshi/nano-banana-2-skillRemote code download and execution detected
curl -fsSL https://bun.sh/install | bash
The skill's initialization process involves cloning a Git repository and installing dependencies using `bun install`. This executes arbitrary code from external sources, posing a significant supply chain risk if the repository or its dependencies are compromised.
git clone https://github.com/kingbootoshi/nano-banana-2-skill.git ... && bun install
The skill explicitly uses `FFmpeg` and `ImageMagick` for transparent asset generation. If the `nano-banana` CLI constructs commands for these tools using unsanitized user input (e.g., from prompts or filenames), it could lead to arbitrary command execution.
The `-t` flag ... uses FFmpeg `colorkey` + `despill` ... Requires: `brew install ffmpeg imagemagick`
The `-d, --dir` option allows specifying an arbitrary output directory. If the CLI does not properly sanitize or restrict this path, it could be abused to write files to sensitive system locations or overwrite critical files.
| `-d, --dir` | current directory | Output directory |
The fallback initialization instructions modify the user's `.zshrc` file to add `~/.local/bin` to the PATH. While common for CLI tools, this is a form of persistence by altering shell configuration.
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc
The skill instructs users to store their Gemini API key in a `.env` file and allows direct input via a flag. While standard practice, it involves handling a sensitive credential, which could be exposed if not managed carefully (e.g., in shell history via the flag).
echo "GEMINI_API_KEY=<ask user for their key>" > ~/.nano-banana/.env ... `--api-key` flag
[](https://mondoo.com/ai-agent-security/skills/github/kingbootoshi/nano-banana-2-skill/nano-banana)<a href="https://mondoo.com/ai-agent-security/skills/github/kingbootoshi/nano-banana-2-skill/nano-banana"><img src="https://mondoo.com/ai-agent-security/api/badge/github/kingbootoshi/nano-banana-2-skill/nano-banana.svg" alt="Mondoo Skill Check" /></a>https://mondoo.com/ai-agent-security/api/badge/github/kingbootoshi/nano-banana-2-skill/nano-banana.svgSkills can read files, run commands, and access credentials. Mondoo helps organizations manage the security risks of AI agent skills across their entire fleet.