Nano Banana 2

Generate images with Google Gemini 3.1 Flash Image Preview (Nano Banana 2) via inference.sh CLI. Capabilities: text-to-image, image editing, multi-image inpu...

This skill appears to do what it says (use inference.sh to run Gemini image apps) but it asks you to download and run a third‑party installer and sign in to an external service. Before installing: (1) Inspect https://cli.inference.sh manually — don't run curl | sh without reading the script. (2) Prefer downloading the binary and verifying SHA-256 checksums yourself using the provided checksums URL. (3) Understand what credentials 'infsh login' will store and where (API tokens, config files). (4) If you have concerns, run the installer in an isolated environment (VM/container) or skip installing and use alternative vetted image-generation tooling. If you want a firmer assessment, provide the actual install script (cli.inference.sh) and any documentation about the login/auth flow so we can inspect exactly what gets executed and what credentials are created/stored.

Type: OpenClaw Skill Name: nano-banana-2 Version: 0.1.1 The skill's purpose is to generate images using the `inference.sh` CLI, which is clearly stated and consistently demonstrated across all examples in `SKILL.md`. The `allowed-tools` permission is appropriately scoped to `Bash(infsh *)`, meaning the AI agent can only execute commands starting with `infsh`. While the `SKILL.md` includes `curl -fsSL https://cli.inference.sh | sh` as an installation instruction for the `infsh` CLI, this is a prerequisite for the human user, not a command the AI agent is allowed or instructed to execute as part of its skill operation. There is no evidence of prompt injection against the agent, data exfiltration, persistence mechanisms, or other malicious intent.