跳到正文
cccskills
免费在 GitHub 公开

commonly-design

Use this skill for ANY visual / UI / design work on Commonly — picking colors, choosing fonts, styling new v2 components, building HTML mocks, designing marketing pages, or auditing whether a change matches the brand. Triggers on tasks involving v2.css tokens, accent / blue / focus ring / SF font, brand mark / logo, design system, design tokens, agent avatars / role tints, pod / chat / inspector / composer styling, marketing landing or login pages, screenshots / demo polish, "make this look better", "design this", "polish UI", "match the brand", "what's the right color/font/spacing/radius". Sources of truth live at `frontend/design-system/` (tokens.css, README.md, brand mark SVG, preview cards) and `frontend/src/v2/v2.css` (production tokens). Read both before recommending or shipping anything.

中文简介正在准备中。当前显示原文说明。

实测评分 未实测

查看安装方法

包含的文件(1)

  • SKILL.md7.8 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Commonly Design

This skill is a thin pointer — the actual design system lives in the commonly repo at frontend/design-system/. Always read those files when invoked, then apply.

Workflow

  1. Read first — every time this skill triggers:
    • docs/design/signal-identity.md — the identity as a system (Signal, chosen 2026-09-03): five rules, the complete grey ramp, state grammar, component grammar with exact values, copy rules, the process that made it land, how to add a screen. Read this FIRST for anything visual.
    • frontend/design-system/README.md — voice, content rules, visual fundamentals, iconography, layout, animation; § Identity: Signal is the short form of the doc above
    • frontend/design-system/tokens.css — the canonical token set as portable CSS
    • frontend/src/v2/v2.css (top of file, the .v2-root { ... } block) — the production tokens that ship today
  2. Match tokens to use case — pick the smallest set of tokens that fit. Don't introduce new colors, sizes, or radii without strong reason; the system is intentionally tight.
  3. Apply through var(--v2-*) — never hardcode hex values in v2 components. If a hex appears in a diff, that's almost always a bug.
  4. Cross-check brand voice — read the "Content fundamentals" section in the README. Sentence case, no emoji, em-dashes, agents-by-name.

The two sources of truth and how they relate

FileWhat it isWho reads it
frontend/src/v2/v2.cssProduction tokens — what actually ships, scoped under .v2-rootThe browser
frontend/design-system/tokens.cssSame tokens as portable --c-* CSS — for HTML mocks, design-review pages, cross-tool skill useDesigners, mocks, this skill
frontend/design-system/README.mdThe system's spec — voice, visual fundamentals, layout (incl. the .v2-root app-shell scroll-containment rule), iconographyHumans + agents during design work
frontend/design-system/preview/*.htmlOne-card-per-concept review pages — open in a browser to see live samplesAnyone reviewing the system
frontend/design-system/assets/commonly-mark.svgBrand mark — uses currentColor so it inherits the surfaceV2NavRail, marketing, login

Hard rule: when one moves, the other moves in the same PR. If you change a token in v2.css, change it in tokens.css too (and vice versa). If they drift, v2.css is what the user sees but the system has lost its source of truth — fix it immediately.

What the system says (read the README for the full version)

One colour, two volumes. Cobalt #1d3fd1 is the only accent. Front door (landing, connect, create account, invite): cobalt as BLOCKS. Inside the app: cobalt as MARKS — live dot, agent name in mono, link, focus edge, the 2px ring on the card that needs you. Per docs/design/signal-identity.md §1–3, an open decision card's primary option is cobalt; secondary options are bordered and Other… is cobalt text. Ordinary filled actions use ink #101828; this does not make every action a filled button. Match the screen artboard's control hierarchy (Activity Reply/Open are bordered).

Three faces, one job each. Bricolage Grotesque 700/800 for display (−0.03em), IBM Plex Sans 400/500/600 for body and controls (14/20), IBM Plex Mono 500 for meta (ids, times, counts, status, commands). Self-hosted via @fontsource; the pre-Signal SF-first stack is gone.

Hard edges, no shadows. Radius 4 on controls, rows, chips, avatars (4px squares, never circles); 6 on cards and panels. No shadow anywhere; a pending card is a 2px cobalt ring, not a lift.

No emoji in chrome. Emoji are user-generated content only.

No gradients in chrome. Solid surfaces, only.

Tight letter-spacing on big text. Feature titles -0.03em at weight 850. Body stays at 0.

Sentence case for buttons, nav, section titles. UPPERCASE only for kickers (NOW, LEAD, CURRENT TASK).

Calm motion. 80ms hover, 120ms card hover, 300ms layout. No bounces, no springs, no entrance animations.

Before you draw or build a screen (the process that made Signal land)

  1. Canvas first. The artboard exists and Sam has seen it before any component work starts. Copy the nearest artboard, change only the content, keep the helmet block.
  2. Directions, when a direction is open: three named candidates, the same product moment, the same fixed copy, so the only variable is the design. Then the chosen one gets clickable on the screen people live in.
  3. One system, two volumes. A new screen never needs a new colour, radius or face. If it seems to, the screen is wrong.
  4. Ship with proof. The PR carries 1440 and 390 screenshots of the real screen beside its artboard; load-bearing CSS is re-pinned in v2-layout-invariants.test.ts.
  5. Survey what is left every few weeks: screenshot every live route into one sheet and give each uncovered screen a call — draw, fold, or delete (Sam rules on deletes).

When you ship a change

  1. Tokens — update frontend/src/v2/v2.css + frontend/design-system/tokens.css in the same commit.
  2. Brand asset — update both frontend/design-system/assets/ and any in-component reference (e.g. inline SVG in V2NavRail.tsx).
  3. README rule change — if you're deviating from a documented rule, update the README in the same PR. The README is the spec; if the spec is outdated, fix the spec.
  4. Preview card — if the change is shape-altering (a new component pattern, a new color role), add or update a preview HTML card under frontend/design-system/preview/.

Improving the system

The design system is meant to evolve. The way to evolve it:

  • Propose changes through the same files. Don't keep design rules in your head or in scratch docs.
  • Document the why in the README when adding a token or rule, not just the what. Someone in 6 months should understand the trade-off.
  • Kill old patterns explicitly. When a token is replaced, remove it from both files. Don't leave deprecated values around — agents will pick them up.
  • Preview before merge. For any visual change, open the relevant frontend/design-system/preview/*.html in a browser to see it in isolation, plus a v2 page in the app to see it in context.

When in doubt

  • For mocks / throwaway prototypes, use frontend/design-system/tokens.css + lucide icons via CDN. Output static HTML the user can open.
  • For production v2 code, use var(--v2-*) tokens from v2.css. Match the existing component patterns (V2PodChat, V2PodInspector, V2MessageBubble).
  • For new component types, draft a preview card first under frontend/design-system/preview/, then implement against tokens.

Related skills

  • frontend-design — Anthropic's public design-judgment skill. Pull it alongside this skill for any new surface or redesign work; this skill carries Commonly's tokens and rules, while frontend-design carries the process that stops templated output.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

commonly

免费

You are a member of a Commonly workspace — a shared space where humans and AI agents from any origin collaborate in pods (chat rooms with memory). Use this whenever you are connected to Commonly via the commonly_* MCP tools: to read what's happening, post, remember things across sessions, react, DM other agents, and work the task board. Load it the moment you see any commonly_* tool available.

中文简介正在准备中。当前显示原文说明。

Team-Commonly/commonly13702026年10月25日 更新

commonly

免费

Use a Commonly pod as shared context for coding work, and post concise handoffs or files to teammates through its MCP tools. Load whenever commonly_* MCP tools are available.

中文简介正在准备中。当前显示原文说明。

Team-Commonly/commonly13702026年10月25日 更新

github

免费

Interact with GitHub (issues, PRs, repos, releases) using the `gh` CLI. Use when asked to read or write GitHub state — open an issue, fetch PR diff, comment, list runs, etc.

中文简介正在准备中。当前显示原文说明。

Team-Commonly/commonly13702026年10月25日 更新

Convert binary documents (PDF, DOCX, XLSX, PPTX, HTML, EPUB, images) to clean LLM-friendly Markdown using Microsoft's `markitdown` Python tool. Use when a user attaches a binary file and you need to read its contents.

中文简介正在准备中。当前显示原文说明。

Team-Commonly/commonly13702026年10月25日 更新

morph-ppt

免费

Use this skill when the user wants a .pptx with smooth cross-slide animation — PowerPoint Morph transitions, Keynote-style continuous motion, shapes that grow / move / rotate as the slide advances. Trigger on: 'morph', 'morph transition', 'smooth transition', 'continuous animation across slides', 'Keynote-style transition', 'animated slide sequence', 'shape continuity across slides'. Output is a single .pptx. This skill is a scene layer on top of officecli-pptx — inherits every pptx v2 rule (visual floor, grid, palettes, connector canon, Delivery Gate 1–5a). DO NOT invoke for a generic deck, pitch deck, or board review without cross-slide motion — route those to officecli-pptx base or officecli-pitch-deck.

中文简介正在准备中。当前显示原文说明。

Team-Commonly/commonly13702026年10月25日 更新

3D Morph PPT — extends morph-ppt with GLB model insertion, cinematographic camera, model-content layout, and enriched visual design system.

中文简介正在准备中。当前显示原文说明。

Team-Commonly/commonly13702026年10月25日 更新

查看 Team-Commonly 的全部技能

报告此技能的问题