Turn a model's official docs into a Claude skill

I've read the official Nano-Banana 2 prompting documentation you attached and built a skill that writes prompts the way this model actually wants them. Here's the skill definition, ready to install:

**Skill: `nano-banana-2-prompter`**

**Triggers:** "nano-banana prompt," "image prompt for nano-banana," or when I explicitly ask for a Nano-Banana 2 generation prompt.

**What it does**
- Writes in Nano-Banana 2's preferred clause order: *subject → scene → action → composition → lighting → palette → style → avoid-list*, comma-chained, no narrative sentences (the docs specify this model parses comma-separated descriptors, not prose).
- Uses the model's native vocabulary: "seamless backdrop," "key/fill/rim," "macro detail," "shallow DOF" — not generic terms like "nice lighting."
- Flags unsupported features before generating: per the docs, Nano-Banana 2 does **not** reliably render legible embedded text or precise hand poses, so it warns me and suggests adding text in post.
- Always asks me for the eight required elements before writing: subject, scene, action, composition, lighting, palette, style, and avoid-list. If I skip one, it picks a sensible default and flags the assumption.

**Front-matter** includes name, description, and the WebFetch-free, self-contained prompt body (no external calls needed).

Install this into `.claude/skills/`, then test with: "Write a Nano-Banana 2 prompt for a matte amber serum bottle." It'll interview you for the missing elements, then output a model-ready prompt plus a one-line rationale per descriptor.