Bonsai-27B-GGUF
lmstudio-community · Kit
A GGUF conversion of the 27B Bonsai model for LM Studio, with a 3.80 GB Q1_0 language file and a 0.93 GB BF16 vision projector. Tested for text, vision and par…

ehmo’s eight design skills for coding agents: iPhone, iPad, Mac, Apple Watch, Vision Pro, Apple TV, Android and the web. They hold rules from Apple’s HIG, Material Design 3 and WCAG 2.2, with code, for reviewing or building screens.
Platform Design Skills is an open-source project on GitHub in the Developer category, listed by Mindket. It is free. This is version 2, last updated on . Install it from GitHub as its README says. Get Mindket’s setup notes on your Mindket account, as a zip.
npx skills add ehmo/platform-design-skills
And 1 more in Where to use it
Platform Design Skills is ehmo’s set of eight skills that teach a coding agent design rules. Each covers one platform: iPhone, iPad, Mac, Apple Watch, Vision Pro, Apple TV, Android or the web. The rules come from Apple’s Human Interface Guidelines, Material Design 3 and WCAG 2.2. Many come with code in that platform’s language. The skills are plain text.
Developers and designers who build apps or sites with an AI agent and want the result to follow each platform’s own conventions. The README says to use the skills to evaluate, improve or create designs. They do not test your app or look at a screen.
| Who | Skills they reach for | Example |
|---|---|---|
| iPhone and iPad developers | ios, ipados | A SwiftUI developer asks whether a screen keeps the 44 point minimum tap target and keeps its content inside the safe areas. |
| Mac and Watch developers | macos, watchos | A developer checks a Mac app’s menu bar and keyboard shortcuts, or a watch screen’s glanceable layout. |
| Android developers | android | A Compose developer checks that a theme uses dynamic color and that the navigation pattern fits the screen size. |
| Web developers | web | A developer asks for an accessibility audit of a page, such as semantic elements, visible focus and labels on form fields. |
Any agent that reads Agent Skills and that the skills installer supports. The maker’s README names no agent, and the installer’s own docs list them.
Get it saves a zip of Mindket’s setup notes in your downloads folder, with the install line, each agent’s folders and a first-run guide.
| Need | Detail |
|---|---|
| An agent | Claude Code, Codex, Cursor, Gemini CLI or another that reads Agent Skills |
| Node.js 22.20 or newer | The installer’s README asks for it, and the install line runs through npx |
| Accounts and keys | None for the pack |
npx skills add ehmo/platform-design-skills
| Option | What it does |
|---|---|
| -g | Installs for all your projects instead of this one |
| -a claude-code | Targets one agent without asking |
| --skill name | Installs one skill by its name |
| --list | Lists the skills in the repository without installing |
| --copy | Copies files instead of linking them |
| -y | Skips the confirmation questions |
Name the platform and what you want checked. The README’s own prompts are below, and a skill’s description tells your agent when it applies.
| Your request | Skill that applies | What Claude does |
|---|---|---|
| Review this SwiftUI view for iOS HIG compliance | ios-design-guidelines | Compares the view with the layout, navigation, type, color and accessibility rules |
| Check this Android Compose screen against Material Design | android-design-guidelines | Compares the screen with the theming, navigation, layout and component rules |
| Audit this web page for accessibility | web-design-guidelines | Compares the page with the WCAG 2.2 based rules and its checklist |
You built a settings screen in SwiftUI and want to know whether it follows Apple’s rules for iPhone.
Review this SwiftUI view for iOS HIG compliance
Claude reads the iOS skill, compares the view with rules such as the 44 point tap target, the safe areas and Dynamic Type, and lists each rule the code breaks.
If you are new. A skill is a text file of instructions that your agent reads when your request matches it. These eight are rulebooks, so they change nothing on their own. Ask for a review first and read the list. Then ask Claude to fix one rule at a time.
For experts. Each skill folder holds a SKILL.md with the rules, a metadata.json, an AGENTS.md of quick context and a rules folder with an index. The installer recommends linking every agent to one canonical copy, and --copy makes separate ones. Unused skills still add a description line to the agent’s list, so install only what you need. The installer’s own commands manage them afterwards:
npx skills list npx skills check npx skills update npx skills remove
The pack is text only.
# Platform Design Skills: setup notes This kit holds setup notes written by Mindket. It does not contain the eight skills, their rule indexes or the Apple HIG PDF that the repository also holds. Those install from ehmo’s repository: https://github.com/ehmo/platform-design-skills License: MIT. The repository’s LICENSE names no copyright holder after the year 2026. Full text: https://github.com/ehmo/platform-design-skills/blob/main/LICENSE Platform Design Skills is a community project by ehmo. It is not made or endorsed by Apple, Google, W3C or Anthropic. It distills their public guidelines into rules for coding agents, and the guidelines stay theirs. The pages these notes came from: the README (https://github.com/ehmo/platform-design-skills/blob/main/README.md), the repository’s AGENTS.md and CHANGELOG.md, and the eight SKILL.md files under https://github.com/ehmo/platform-design-skills/tree/main/skills. The installer’s facts come from the README of the skills command line it names, Vercel Labs’ skills repository (https://github.com/vercel-labs/skills), and from its docs/commands.md and docs/agents.md. Facts about where Claude Code loads skills come from Claude Code’s own skills page (https://code.claude.com/docs/en/skills). Mindket copied every command and ran none of them. ## What it is Eight skills for coding agents, one per platform. Each holds numbered or coded rules with code in that platform’s language, so an agent can check a screen or write one against them. The README says there are 450+ rules in all. | Skill name | Platform | Guideline behind it | |---|---|---| | ios-design-guidelines | iPhone | Apple Human Interface Guidelines | | ipados-design-guidelines | iPad | Apple Human Interface Guidelines | | macos-design-guidelines | Mac | Apple Human Interface Guidelines | | watchos-design-guidelines | Apple Watch | Apple Human Interface Guidelines | | visionos-design-guidelines | Apple Vision Pro | Apple Human Interface Guidelines | | tvos-design-guidelines | Apple TV | Apple Human Interface Guidelines | | android-design-guidelines | Android | Material Design 3 | | web-design-guidelines | Web | WCAG 2.2 and MDN | Source: the repository’s AGENTS.md table, and the name line at the top of each SKILL.md. The README calls the folders ios, ipados, macos, watchos, visionos, tvos, android and web. ## Who does what - You: run the install line, then ask Claude to build or review a screen in plain words. - The installer: puts the skill folders where your agent reads them, as links to one shared copy or as separate copies. It is a separate program, not the maker’s. - Claude (or another agent): reads a skill when your request matches its description, then checks your code against the rules or writes new code that follows them. The rules are text, so nothing in the pack runs code or checks a screen for you. ## Files in this kit - setup/skills-cli.md: the README’s one install line, and what the installer asks and how to pick skills, scope and agents - setup/claude-code.md: installing for Claude Code, where the skills land, and checking them - setup/other-agents.md: Codex, Cursor, Gemini CLI, GitHub Copilot and the other agents the installer names - setup/first-run.md: the README’s example prompts, what each skill covers, and what to double check ## What you need first - An agent that reads Agent Skills, such as Claude Code, Codex, Cursor or Gemini CLI. - Node.js 22.20 or newer, since the installer’s README says so. The install line runs through npx. - A project in git or a plain folder, if you want a project-only install. A global install needs none. - No account, no API key and no other software for the pack. ## What installs where - Project install, the installer’s default: the skill folders go in your project, in the agent’s own skills folder. For Claude Code that is .claude/skills/. - Global install, with the -g option: the folders go in your home directory. For Claude Code that is ~/.claude/skills/. - Each skill folder holds a SKILL.md with the rules, a metadata.json, an AGENTS.md and a rules folder. The repository’s AGENTS.md says that folder holds _sections.md, an index of the rules by category. ## Your data The skills are text. They tell the agent what the guidelines say and send nothing anywhere. The installer downloads the repository from GitHub. Its README says it collects anonymous usage data, and that setting DISABLE_TELEMETRY=1 or DO_NOT_TRACK=1 turns that off. What you give Claude, such as your code and screenshots, goes to Claude as in any other Claude work. ## Update and remove The maker’s README gives no update or removal step. The installer’s README gives npx skills update and npx skills remove, shown in setup/skills-cli.md.
README.md 5 KB setup/claude-code.md 3 KB setup/first-run.md 4 KB setup/other-agents.md 2 KB setup/skills-cli.md 3 KB
Updated by Mindket to the current version of this kit.
Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search and on Trendshift. Mindket doesn’t write or check them. Open one to read it where it was posted.
lmstudio-community · Kit
A GGUF conversion of the 27B Bonsai model for LM Studio, with a 3.80 GB Q1_0 language file and a 0.93 GB BF16 vision projector. Tested for text, vision and par…
LibertAIDAI · Kit
Runs DeepSeek-V4.1-Flash-1017 as a 2.9 bits-per-weight EXL3 quantization, a 196 GiB copy for people serving large models on their own hardware. Perplexity stay…
facebook · Kit
Generates short music clips from a text description or an audio prompt, in the small checkpoint of facebook's MusicGen. It suits developers and music hobbyists…
alpindale · Kit
Tags every character of C source code for syntax highlighting with a tiny neural network of about 984 weights, run through a small C program. For developers an…
prithivMLmods · Kit
Try a set of LoRAs for the FLUX.2 Klein image model in a Gradio app that runs in your browser or on your own computer. Made for people who make or test AI imag…
Sokheng · Kit
Reads Khmer and English text from documents, receipts and photos of signs, using a detector for text lines and a recogniser for each line. Suits developers who…