bama
BAMA: Build Any Mac App. Agent skills to research, clone, design, build, optimize, and test native macOS apps with Swift and SwiftUI.
翻譯準備中
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證Agent skill for studio-grade web design, measured from 99 live sites by top studios. Self-audits its own work.
翻譯準備中
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
An agent skill for studio-grade web design. It was measured from 99 live sites built by about 40 studios and in-house teams: basement.studio, Exo Ape, Work & Co, AREA 17, Locomotive, Lusion, Malvah and others. Your agent picks a direction, sets how loud to be, builds from measured tokens, and audits its own work before you see it.
This works with Claude Code, Cursor, Codex, Windsurf, Gemini, Copilot and the other agents that support skills.sh. The self-audit needs Playwright in the project you're building: npm i -D playwright && npx playwright install chromium.檢視完整 README ↗▶ Watch the launch film (35s), or download the MP4.
An agent skill for studio-grade web design. It was measured from 99 live sites built by about 40 studios and in-house teams: basement.studio, Exo Ape, Work & Co, AREA 17, Locomotive, Lusion, Malvah and others. Your agent picks a direction, sets how loud to be, builds from measured tokens, and audits its own work before you see it.
npx skills add arjunkshah12345-hash/studio-grade
This works with Claude Code, Cursor, Codex, Windsurf, Gemini, Copilot and the other agents that support skills.sh. The self-audit needs Playwright in the project you're building: npm i -D playwright && npx playwright install chromium.
Site and work: studio-grade.vercel.app. It has sixteen sites built by agents that read only this skill, each with its Design Spine, palette, every screen and its audit result.
node studio-grade/scripts/audit.mjs <page> --direction N --register R checks 30+ rules at 1440 and 390px, flags the default AI look, and saves a screenshot of every screen for the agent to look at.The agent runs scripts/audit.mjs on its own page. Here a template page fails it, and a page built with the skill passes. The output is real, abridged.
Each rule exists because a real build needed it. One build's headings stayed blurred because a CSS selector outranked the reveal state, so motion.stuckReveal now fails it. Sticky headlines had text scroll over them, so layout.textOverlap now fails that too.
| Path | What it is |
|---|---|
studio-grade/ | The skill: SKILL.md, 12 references, assets (tokens.css, motion.js), scripts/audit.mjs. This is the only folder that gets installed. |
tests/builds/ | The gallery sites, built by agents that read only the skill (most with a generated image series). |
tests/feedback/ | Each test builder's notes on the skill. They drove the fixes. |
research/ | Per-cluster notes and the aggregate of the 99-site teardown. |
web/src/, tools/web/ | The website, built with the skill. shoot.mjs captures the work, and build.mjs writes web/dist/. |
cd tools && npm i && cd ..
python3 -m http.server 4317 & # serves tests/builds for shoot.mjs
node tools/web/shoot.mjs # optional: re-capture screens into web/src/media
node tools/web/build.mjs # → web/dist
node studio-grade/scripts/audit.mjs http://127.0.0.1:4318/ --register confident # serve web/dist on 4318
The sites studied belong to their owners and the studios that made them. We measured public pages; none of their work is reproduced here.
MIT © Arjun Shah