design-system-author
How to build a Figma Foundation page (design tokens + reusable components) BEFORE drawing screens, so spacing / text sizes / button styles / element alignment…
Skill by evolplus
Chalk Labs
Need help deploying design-system-author?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/evolplus-design-system-author
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design, security
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/design-system-author
What's inside
- Skill name
design-system-author- Description
- How to build a Figma Foundation page (design tokens + reusable components) BEFORE drawing screens, so spacing / text sizes / button styles / element alignment stay consistent across every surface. Consult on every UI/UX Designer dispatch (create / revise / import / incorporate). Pre-handoff token-compliance lint mandatory in every mode.
- License
- Apache-2.0
- Instructions
- 498 lines in SKILL.md · view on GitHub
Files (12)
- SKILL.md
- README.md · references/presets/
- components.md · references/presets/_template/
- layout-grid.md · references/presets/_template/
- preset.md · references/presets/_template/
- tokens.json · references/presets/_template/
- typography.md · references/presets/_template/
- components.md · references/presets/default/
- layout-grid.md · references/presets/default/
- preset.md · references/presets/default/
- tokens.json · references/presets/default/
- typography.md · references/presets/default/
SKILL.md (first lines)
# Design System Author ## When to use You are the UI/UX Designer authoring or revising Figma content via MCP. Without a shared Foundation, every screen invents its own spacing scale (16px here, 12px there), text sizes (18pt vs 20pt for the same role), button radii (8px vs 12px), and alignment grid — drift compounds across screens and the Approver sees visual chaos at confirmation time. This skill is the **content-hygiene** complement to [`figma-canvas-layout`](../figma-canvas-layout/SKILL.md) (canvas-hygiene): same pre-handoff lint pattern, different concern. Consult this skill: - **In every Figma-writing mode** (`create` / `revise` / `incorporate`) — apply the Foundation tokens, never invent ad-hoc values. - **In `import` mode** — extract what tokens already exist in the file; flag gaps as `figma-design-system-gap` open-issues. - **As a pre-handoff lint** — every mode, every dispatch. ## Inputs and outputs - **Inputs:** SRS §3.4.1 Design References (UI surfaces), SRS `## Design System` or brand-guidance section if present, the Figma file you're writing to via MCP. - **Outputs:** A `Foundation` Figma page (separate Figma page tab) containing tokens + components; every other page's frames consume from Foundation; an updated `docs/uiux/handoffs/<task-id>.md` enumerating tokens used; a pre-handoff token-compliance lint report appended to the handoff doc. ## The Foundation page Every Figma file the kit produces has a dedicated **Foundation** Figma page (separate page tab, NOT a section on a screens page). The Foundation page is the single source of truth for tokens + components. Screen pages reference it; screens never define their own colors, type styles, or spacing values. Foundation is structured as five Figma sections. **Before authoring those sections from scratch**, run the Foundation-source selection step below — most projects inherit from a named preset, while Flow A projects may derive the Foundation from extracted Figma evidence. ### Step 0 — Foundation source selection (run BEFORE any of Sections 1-5) The kit ships a catalog of named presets at `references/presets/<slug>/`. Each preset is a pre-authored Foundation: tokens, type scale, component minimum set, layout grid. For Design-Flow A, the SRS may instead declare `Design-Guideline: from-figma`, meaning the existing Figma file itself is the Foundation source and the Designer must build or audit against the design-extracted token evidence. For every Figma-backed dispatch, the UI/UX Designer still captures token evidence from the Figma link regardless of the declared `Design-Guideline:` value. A preset or `none` value decides the Foundation source; it does not waive extraction of the live Figma palette, typography, spacing, radius, elevation, component-pattern, and grid evidence. The handoff must make clear which values are preset/defaulted and which values came from Figma. #### Procedure 1. **Read the SRS header.** `docs/SRS.md` declares the chosen Foundation source in the `Design-Guideline:` field (set by BA Phase 1.X step 10b at sign-off time). Possible values: - A preset slug → e.g., `modern-saas-admin`, `ios-consumer`, `default`. - `from-figma` → Design-Flow A extracted enough design-token evidence from the provided Figma file to use it as the Foundation source. Build/audit Foundation from `docs/requirements/design-extracted/<figma-file-id>-*.md` Section 6. - `none` → no preset; author Foundation from the SKILL.md defaults (Sections 1–5 below). - Missing field → halt with `NEEDS_CONTEXT` and ask the Orchestrator to dispatch BA Mode D to set the field. Do NOT pick a preset unilaterally; the choice is BA + operator territory. 2. **For every Figma URL, load or require token evidence.** Read the latest `docs/requirements/design-extracted/<figma-file-id>-*.md` for each Figma URL in SRS §3.4.1 and inspect `## Section 6 — Design guideline extraction (Flow A)`. If no extraction exists for a Figma-backed project, halt and ask the Orchestrator to dispatch UI/UX Designer in `extract` mode before design handoff/implementation proceeds. - This applies even when `Design-Guideline:` is a preset slug or `none`; those choices do not erase the need to inventory what the Figma file actually uses. - If the live Figma file version is newer than the extraction evidence, halt with `NEEDS_CONTEXT` and ask whether to re-run extraction. 3. **If `Design-Guideline: from-figma`, use extraction evidence as the Foundation source.** - Use formal Figma styles/variables/components first when Section 6 identifies them. - When no formal styles exist, use medium/high-confidence repeated values from Section 6 as token candidates. - Build Sections 1–5 from the extracted palette, typography, spacing, radius, elevation/effects, component patterns, and layout grid evidence. - Preserve numeric values from the extraction. Do not "round" a 6px radius to 8px or convert a 20px spacing value to 24px unless the handoff explicitly records it as a fallback/deviation. - If a required token category is absent, fill the missing category from SKILL.md defaults and record the fallback in `## Design System Source` and `## Foundation Changes`. - If Section 6 is missing, low-confidence, or contradictory, halt with `NEEDS_CONTEXT`; do not silently switch to a preset. 4. **If a preset slug is declared, load the preset.** Read every file under `.claude/skills/design-system-author/references/presets/<slug>/`: - `preset.md` — when-to-use, dos / don'ts (apply throughout your authoring). - `tokens.json` — color + spacing + radius + elevation + motion + z-index tokens. **Apply these values verbatim to the Figma Foundation page's Sections 1, 3, 4 — do NOT invent your own values.** - `typography.md` — type scale + font stack + rules. **Apply to Section 2.** - `components.md` — required component set + variants + states + preset-specific additions. **Apply to Section 5.** - `layout-grid.md` — column system + breakpoints + container widths + vertical rhythm rules. **Apply to your page-level layouts and document in `## Foundation Changes` if you need to deviate.** 5. **Honour source-specific rules.** `preset.md` carries Dos / Don'ts for preset sources; Section 6 carries source evidence and gaps for `from-figma`. These ARE Foundation rules — bake them into your component variants + screen authoring. Example: a preset may forbid `text-transform: uppercase`; a `from-figma` extraction may show all cards use 12px radius and 24px internal padding. Both are enforced via the lint at Step 3 below. 6. **Note deviations explicitly.** If the SRS implies a Foundation value that conflicts with the declared Foundation source (e.g., SRS §3.4.1 specifies a custom brand color), document the deviation in the handoff doc's `## Foundation Changes` section. Do NOT silently override the source; the explicit annotation lets FE Dev + QA-Author know which values are preset-stock, Figma-extracted, or project-customized. #### When to clone a preset If the project consistently diverges from its declared preset on >3 token values, the cleaner long-term move is to clone the preset to a new slug: 1. `cp -r references/presets/<parent>/ references/presets/<new-slug>/`. 2. Edit the new preset's `preset.md` § Provenance to cite the parent + diff summary. 3. Apply the changes only to the new preset. 4. Re-dispatch BA Mode D to update SRS `Design-Guideline:` to the new slug. 5. Append the new preset to `.claude/skills/registry.md` § Design Guidelines. Cloning preserves the parent for other projects and lets this project diverge cleanly. Avoid editing the shared preset in place for project-specific reasons. #### When the preset is `none` If `Design-Guideline: none`, author Foundation from the SKILL.md defaults below (Sections 1–5). Document in the handoff that no preset applies — FE Dev + QA-Author then derive tokens from your authored Foundation directly, not from any preset path. #### When the guideline is `from-figma` If `Design-Guideline: from-figma`, the Flow A design-extracted file is the source of truth. Build or audit the Foundation from its Section 6 token evidence: - Colors, typography, spacing, radius, elevation/effects, component patterns, and layout grid values come from the extracted Figma evidence. - The handoff `## Design System Source` section must cite the exact extraction artifact path, Figma version, and confidence level for every token category. - Missing token categories use SKILL.md defaults only as explicit fallbacks. Each fallback is recorded so FE Dev and QA-Author know which values are extracted vs synthesized. - If the live Figma file has drifted from the extracted evidence, halt with `NEEDS_CONTEXT` and ask whether to re-run `figma-requirements-extraction` before authoring. --- ### Section 1 — Color tokens Define named colors as Figma color styles. Two layers: - **Primitives** — raw color values (e.g., `gray-50`, `gray-100`, ..., `gray-900`; `blue-500`; `red-500`). Don't use primitives directly in screens. - **Semantic** — purpose-bound aliases referencing primitives (e.g., `text-primary` → `gray-900`; `text-secondary` → `gray-600`; `surface-default` → `white`; `accent-brand` → `blue-500`; `state-error` → `red-500`; `state-success` → `green-500`; `border-default` → `gray-200`). Screens consume semantic tokens only. Why two layers: changing the brand color updates `accent-brand`'s primitive reference once; every screen that consumes `accent-brand` updates automatically. Direct-primitive consumption defeats this. ### Section 2 — Typography tokens Define named text styles as Figma text styles. Default scale (modular ratio 1.250, 16px base): | Name | Size | Weight | Line height | Use | |---|---|---|---|---| | `heading-xl` | 48px | 700 | 56px | Page-hero title | | `heading-lg` | 32px | 700 | 40px | Section title | | `heading-md` | 24px | 600 | 32px | Card title | | `heading-sm` | 20px | 600 | 28px | Inline group title | | `body-lg` | 18px | 400 | 28px | Lead paragraph | | `body-md` | 16px | 400 | 24px | Default body text | | `body-sm` | 14px | 400 | 20px | Secondary text, captions | | `label-md` | 14px | 600 | 20px | Form label, button text | | `label-sm` | 12px | 600 | 16px | Tag, badge, micro-label | | `code-md` | 14px | 400 | 20px | Monospace inline code | Pair with a defined font family (e.g., Inter for body, system monospace for code). Document the family name in Section 2. ### Section 3 — Spacing tokens Define named spacing values as a 4pt baseline scale (override to 8pt if SRS's design system mandates):
From evolplus/talos, Apache-2.0.
More from evolplus/talos
- 0.631score
be-framework-coding-standard
evolplus
Standardized backend implementation guidance for BE Dev using the SRS-declared Backend-Track and Backend-Framework.
SkillAFreshness: A★ 8Claude Code +9 - 0.628score
api-contract-author
evolplus
How to publish, freeze, and version an API contract under docs/api-contracts/.
SkillAFreshness: A★ 8Claude Code +9 - 0.628score
adr-author
evolplus
How to write an Architecture Decision Record under docs/decisions/.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-post-implementation
evolplus
BA Phase 5 — post-implementation completeness verification (UI tasks).
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-design-completeness
evolplus
BA Phase 3 — post-sign-off design completeness verification.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-iteration-planning
evolplus
BA Phase 4 — iteration dispatch planning.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-mode-external-source
evolplus
BA Phase 1.C ingestion mode (ingest-from-external-source).
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-mode-reverse-engineer
evolplus
BA Phase 1.E ingestion mode (reverse-engineer-from-code) — brownfield onboarding Stage 3.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
backend-logging-traceability
evolplus
Backend logging and traceability standard for BE Dev.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
codebase-inventory
evolplus
Produce the deterministic Tier-1 inventory of an existing codebase — routes/RPC/jobs, broker channels, persistent schema, external egress, deployables…
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
data-lifecycle-contracts
evolplus
How SA identifies gate fields and cross-component column-write contracts in the data model, and documents them in docs/architecture.md §6 so BE/FE Dev know…
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-mode-requirements-folder
evolplus
BA Phase 1.F ingestion mode (ingest-from-requirements-folder) — greenfield with fragmented upstream.
SkillAFreshness: A★ 8Claude Code +9
Signals
- Stars
- 8
- Forks
- 0
- Open issues
- 1
- Last commit (this folder)
- Jul 3, 2026 (4 months ago)
- Last release
- Oct 12, 2026 (19 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/evolplus/talos
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from evolplus
- 0.631score
be-framework-coding-standard
evolplus
Standardized backend implementation guidance for BE Dev using the SRS-declared Backend-Track and Backend-Framework.
SkillAFreshness: A★ 8Claude Code +9 - 0.628score
adr-author
evolplus
How to write an Architecture Decision Record under docs/decisions/.
SkillAFreshness: A★ 8Claude Code +9 - 0.628score
api-contract-author
evolplus
How to publish, freeze, and version an API contract under docs/api-contracts/.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-iteration-planning
evolplus
BA Phase 4 — iteration dispatch planning.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-design-completeness
evolplus
BA Phase 3 — post-sign-off design completeness verification.
SkillAFreshness: A★ 8Claude Code +9 - 0.626score
ba-mode-external-source
evolplus
BA Phase 1.C ingestion mode (ingest-from-external-source).
SkillAFreshness: A★ 8Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get design-system-author running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
design-system-author skill. Agent Mart. https://agentmart.dev/skills/evolplus-design-system-author (last updated 2026-10-24)