auto
Scans your installed skill library and recommends the right taste/aesthetic/domain skills before a build task starts.
Open source Open in the app JSON README (API)
About
Scans your installed skill library and recommends the right taste/aesthetic/domain skills before a build task starts.
Details
- Kind
- Plugins
- Topic
- No topic detected
- Publisher
- viraj-rgb
- Origin
- gemini
- Category
- ferramentas
- Version
- 0.1.0
- Stars
- 1
- Last push
- 2026-09-10T08:18:20Z
- Repository state
- ativo
- License
- MIT
- Added
- 2026-09-10 14:02:51
- Updated
- 2026-09-10 14:02:51
- Origin id
viraj-rgb/auto-skill
README
<p align="center"> <strong align="center">Stop shipping generic AI-slop UI. Route every build to the right skill first.</strong> </p> <p align="center"> <a href="LICENSE"><img src="https://img.shields.io/github/license/viraj-rgb/auto-skill?style=flat" alt="License"></a> </p> ## Install Copy/paste into your CLI prompt: ```text Install the auto skill/plugin from https://github.com/viraj-rgb/auto-skill, refer to the repo's AGENTS.md for instructions. ``` Or 🔗 [check the installation instructions](INSTALL.md). ## What it does A skill advisor for your coding assistant. Before it writes a line of code for a website, app, dashboard, or component, it scans your installed skill library, reads what's actually available (taste skills, aesthetic skills, domain skills), and tells you which ones fit — instead of defaulting to the same purple-gradient, emoji-header, cookie-cutter-bento output every agent reaches for when nothing routes it elsewhere. ## What changes <table> <tr> <td width="50%"> ## Before > *User: "build me a landing page for my coffee shop"* > > Agent generates a page with a purple-to-blue gradient hero, an emoji ☕ as the logo, three feature cards in a generic bento grid, and Inter font at default weights — the same shape it would produce for a fintech dashboard or a dev tool. </td> <td width="50%"> ## After > *User: "build me a landing page for my coffee shop"* > > Agent: "For this: landing page, going for warm/cozy. Going with: > - `no-ai-slop` — baseline pass against generic AI patterns > - `cafe` — warm tones, soft typography fits the brief > - `pexels-images` — real photography instead of stock placeholders > > Sound right, or want a different direction?" </td> </tr> </table> ## The rules Five steps. Full text in [SKILL.md](./skills/auto/SKILL.md). 1. Understand the project — type, audience, mood, any explicit style cues. 2. Scan the installed skill library live; don't trust memory of what was there last time. 3. Match against two layers: a taste/quality-gate skill (catches AI-slop) and an aesthetic-direction skill (the actual visual language) — plus a domain skill when the task needs a specific pipeline. 4. Present the shortlist with one-line reasons, then wait for confirmation. 5. Invoke in order — taste first, aesthetic second, domain third — and build. ## Tune it Fork, edit `skills/auto/SKILL.md`, then swap your copy in: ```bash claude plugin uninstall auto # drop the upstream copy first: claude plugin marketplace remove auto # fork and upstream share both names claude plugin marketplace add <your-username>/auto-skill claude plugin install auto@auto ``` Restart Claude Code, then re-invoke `/auto`. ## License MIT. Star ⭐ if it saved your build from another purple gradient.