AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Extract Design System

skill-therapys-dotagents-extract-design-system · by therapys

Reverse-engineer a project's real, shipped design language into a reusable design skill, so any agent can build or restyle pages that match it. Use when the user says 'create a design skill from this project', 'make a design skill from the current project', 'extract the design system', 'capture this app's design as a skill', 'codify our design language', or 'reverse-engineer the design system int…

No reviews yet
0 installs
0 views
view→install

Install

$ agentstack add skill-therapys-dotagents-extract-design-system

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-therapys-dotagents-extract-design-system)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
2d ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Extract Design System? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Extract Design System

Turn a codebase's actual, shipped design language into a reusable skill. The output is a new SKILL.md — a tight rules + workflow + pitfalls guide — that lets any agent restyle or build pages that look like they belong in this app. Ground everything in the real code and the rendered UI. Never invent tokens; a design system you can't point to in the tree is a guess.

This is the generator that produces a design skill. It is not frontend (which invents a new aesthetic) — here the aesthetic already exists and your job is to codify it faithfully.

Steps

  1. Read the tokens — don't guess. Find where design decisions actually live:
  • Config tokens: Tailwind config (theme.extend — colors, fontFamily, borderRadius,

boxShadow, spacing), CSS custom properties (:root, @theme, data-theme), SCSS/LESS variables, or a tokens.* / theme.* file.

  • Typography: font loading (next/font, @font-face, ``), families, weights, the

type scale, and any display/accent face used differently from body.

  • Color & theming: the palette and the semantic names layered on it (background,

surface, muted, accent, border), plus the dark-mode strategy (class vs prefers-color-scheme) and how token pairs flip.

  • Scales: the real radius, spacing, shadow, and border values in use — and which size

maps to which kind of element (control vs card vs sheet).

  • Components: open the base primitives (Button, Card, Input, layout shell). Note

default radii, whether surfaces use borders / shadows / tone, and the variant set.

  • Motion & atmosphere: transition durations/easings, hover/active patterns,

prefers-reduced-motion handling, and any grain, gradient, glow, or background wash.

  1. See it, don't just read it. Run the app (or open Storybook / existing screenshots) and

look at 2–3 real screens in both themes. Code lies — dead tokens, inline overrides, one-off styles. Trust what renders; reconcile it with the tokens.

  1. Derive the rules, not the inventory. A design system is the handful of non-obvious

constraints that make pages look right — what defines a surface, what never gets a border, the radius scale, how sections are separated, the one signature move. Aim for 6–9 load-bearing rules, each citing real class/token names. A dump of every variable is not a design system.

  1. Write the skill. Create skills/-design/SKILL.md following the skeleton below,

named after the product or aesthetic. Put copy-paste token values and 2–4 real code snippets from the app in a sibling REFERENCE.md if they run long; keep SKILL.md dense.

  1. Validate and dogfood. Run npm run check, then apply the new skill to restyle one

small component and screenshot it in both themes — it should be indistinguishable from the source app. If it isn't, a rule is wrong or missing.

Output skeleton (the design skill you generate)

---
name: -design
description: Apply the  "" design system — . Use
  when building or restyling pages/components in , or when the user mentions
  "" or asks to make a page match .
---

#  Design System

## The N rules
1. **.** 
   … (6–9 rules)

## Workflow for a new page
1. … the order you actually build a screen in this system

## Pitfalls (each shipped as a bug once)
-  → 

## Tokens / reference

Model the density and voice on the project's own well-written docs — every line should save the next agent a lookup.

Guardrails

  • Grounded only. Never document a token, class, or component you haven't found in the

tree and confirmed on screen.

  • Rules over inventory. 6–9 constraints that carry the look beat a wall of variables.
  • Portability. An extracted design skill is usually product-specific (brand colors,

product name, private aesthetic). Keep it local unless it's genuinely generic and meant to be shared — add skills/-design/ to .gitignore so brand-specific design skills never publish by accident. Strip any secret, internal URL, or machine-specific path from the output.

  • Leave committing to the commit / ship skills.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.