design-tokens

design-tokens is a skill for Claude Code, Codex from glebis/humane-agentic-design. It costs 186 tokens per session (6,071 once invoked), scanned A, original, MIT.

A toolkit for managing design tokens in the DTCG 2025.10 format. Design tokens are reusable design settings such as colors, sizes, durations, fonts, shadows, and typography, which it can validate, combine, resolve, and export as CSS.

In plain words
What is it for?
It is for setting up token files, merging base and project settings, resolving aliases, measuring text contrast, exporting CSS, and generating previews or prompts.
Why use it?
It reduces repeated design values and catches invalid or insufficiently contrasted color choices across themes.

Skill for Claude CodeCodex

Written for Claude Code and Codex: shipped in a Claude Code plugin, but also agents/openai.yaml present. Also seen: names the AskUserQuestion tool; mentions Claude Code.

Part of the humane plugin — 21 skills, 1 agent shipped together

Good fit It is for setting up token files, merging base and project settings, resolving aliases, measuring text contrast, exporting CSS, and generating previews or prompts.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/glebis/humane-agentic-design/design-tokens
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add glebis/humane-agentic-design --skill design-tokens
Clone the repo
git clone --depth 1 https://github.com/glebis/humane-agentic-design

Made for: Claude Code, Codex.

Or install humane, the plugin that ships this one along with the rest of its 21 skills, 1 agent.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for design-tokens

README.md
[![agentmods](https://agentmods.dev/badge/skills/glebis/humane-agentic-design/design-tokens/github.svg)](https://agentmods.dev/skills/glebis/humane-agentic-design/design-tokens)
Your own site
<a href="https://agentmods.dev/skills/glebis/humane-agentic-design/design-tokens"><img src="https://agentmods.dev/badge/skills/glebis/humane-agentic-design/design-tokens/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for design-tokens

Your own site · 80×15
<a href="https://agentmods.dev/skills/glebis/humane-agentic-design/design-tokens"><img src="https://agentmods.dev/badge/skills/glebis/humane-agentic-design/design-tokens.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 186 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 6,071 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 2 findings, up to high

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • high YARA Match · line 3
    YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
    Fix: Remove offensive tool references and exploit code. Legitimate agent skills should not contain penetration testing tools, exploit frameworks, or reconnaissance utilities.
  • high Prompt Injection · line 167
    Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
    Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00186 $0.06071
Opus 5 $0.00093 $0.03036
Sonnet 5 $0.00037 $0.01214
Haiku 4.5 $0.00019 $0.00607

Measured yesterday against content hash ba9288eece92, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

design-tokens scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

The scan reads SKILL.md. This mod also ships 1 executable file (conftest.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

humane/skills/design-tokens/SKILL.md · 301 lines

How it starts

The opening of the file, as written. The whole thing — 301 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Design Tokens

Announce at start: "I'm using the humane:design-tokens skill to set up, resolve, and export this token set."

Manage DTCG 2025.10 design tokens with a dependency-free Python core. v1 covers the deterministic spine: scaffold, validate, merge (global base + project override), resolve aliases, and export CSS.

Standard vs convention

  • Standard (DTCG): *.tokens.json, $value/$type, whole-value {alias} references.
  • Skill convention (NOT DTCG): global-base / project-override layering via merge, and theme-as-override-file. These are labelled in code; do not present them as standard.

v1 scope

Supported $type: color (string values), dimension, duration, fontFamily, fontWeight, number, typography, shadow. Outputs: CSS custom properties, a Google-Labs DESIGN.md (alpha), a standalone HTML preview, and generation prompts (gpt-image-2 / nano-banana CLI lines + a /tufte-report theme) via the prompt door. CSS import covers color/dimension/duration/fontFamily/number; composite values (box-shadow, gradients, multi-part typography) are skipped and reported. Not in v1: JSON Pointer $ref, $root, structured color objects, name-restriction enforcement, Style Dictionary, Figma/Pencil importers, share bundles, skillify (see the phased spec).

Commands

Run via scripts/tokens <command> (or PYTHONPATH=scripts python3 -m dtokens.cli):

Command What it does
setup-edit <dest> [--from SRC] Scaffold a token file at <dest> and validate it (refuses to overwrite). With --from, deterministically clone an existing set's structure + content to edit (byte-stable for a given source) instead of the blank template. Ships templates/base.tokens.json (minimal), templates/monaspace.tokens.json (a real set extracted from a live site — see Extracting from a site), and templates/gsap.tokens.json (a motion-first set: duration tokens under motion.* render as a body Motion table, and its $extensions brand block carries GSAP animation recipes for --rich). Also compiles a DESIGN.md next to the token file (provenance-stamped; won't clobber a hand-edited one — see DESIGN.md output). If a brand-block.draft.json (a humane:brandkit handoff for a set that didn't exist yet) sits in the destination directory, its $extensions brand block is imported into the scaffolded set.
import <css> [-o OUT] Import a CSS file's :root custom properties into DTCG, preserving variable names. Skips composites (shadow/gradient) and reports them on stderr.
validate <file> [--strict] Print OK or a list of errors; exit 1 if invalid. Also prints non-fatal warning: lines to stderr when the brand-style block / its imageryStyle is missing, or when a dimension/duration value is legitimate CSS but not DTCG-shapeable (clamp()/calc()/var() and bare unit strings — kept verbatim in outputs). Advisory only, never changes the exit code — except under --strict, which promotes the non-DTCG dimension advisories to errors (exit 1) so they can gate CI.
contrast <file> [--standard apca|wcag|both] [--level auto|body|non-body|graphic] [--json] [--no-fail] [-o OUT] Measure APCA Lc and WCAG 2.x ratio for every foreground/background pair in the set, and propose a fix that moves OKLCH lightness while preserving chroma and hue. Exits 1 on any failure (--no-fail to report without gating). See Contrast below.
merge <base> <override> [-o OUT] Layer project override on global base.
resolve <file> [-o OUT] Flatten aliases to concrete values (JSON map).
export-css <file> [--selector SEL] [-o OUT] Emit CSS custom properties.
design-md <file> [--name N] [--description D] [--rich] [--yes] [-o OUT] Emit a Google-Labs DESIGN.md (alpha) — YAML token frontmatter + a table-based body (colors carry their $description as a Role column). It is provenance-stamped (generated, do-not-edit) and, when the $extensions brand block is present, renders a default ## Brand direction section (mood / imageryStyle / subjects / avoid). --rich (also on use) appends style-guide sections from the brand $extensions block — components, do's/don'ts, surfaces, imagery, layout, similar brands, plus a Quick Start CSS block. Non-standard: --rich extends the Labs alpha body, so the CLI shows a confirmation (auto-accepted with --yes or when non-interactive; the note still prints to stderr).
preview <file> [--name N] [--full] [--description D] [-o OUT] Emit a standalone HTML swatch page (colors, type specimens, spacing, rounded, shadow). With --full, emit a landing-page mockup instead — the brand applied in situ (hero, prose, accent band, footer), driven entirely by the role/type/space tokens via :root vars. Type specimens load their families via a deterministic Google Fonts @import so brand faces render (degrades to a generic fallback offline / for non-Google fonts).
prompt <file> [--target gpt-image-2|nano-banana|tufte|all] [--preset P ...] [--platform P] [--subject S] [--name N] [-o OUT] The prompt door: turn resolved tokens into ready-to-paste generation prompts. Image targets emit per-preset CLI invocations with the brand's hex/fonts/shape baked into the subject; tufte emits a CSS :root theme mapping brand roles onto /tufte-report's variables.
use <file> [--name N] [--description D] [--out-dir DIR] [--serve/--no-serve] [--port N] [--no-open] Validate + resolve, then write tokens.css, DESIGN.md, preview.html, preview-full.html (landing-page mockup), image-prompts.md, and tufte-theme.css. Serves the output over HTTP and opens it by default when interactive (see below).
generate <file> [--target gpt-image-2|nano-banana|all] [--subject S] [--refs DIR] [--out-dir D] [--final] [--dry-run] Actually generate on-brand images: composes the winning art-direction-prose prompt (fidelity-tested) from tokens + $extensions brand block and shells out to the gpt-image-2 / nano-banana skill scripts (cheap draft by default; --final = high/pro). With --refs, reads the refs.json manifest: each annotated image becomes a --reference flag plus a role-annotated prompt clause ("from reference image 1 take: palette, mood — …").
annotate <dir> [--port N] [--no-open] Serve a reference-image annotator for a directory of images: per-image role chips (style, palette, composition, subject, texture, typography, mood) + a free-text note, with voice dictation via Groq Whisper when GROQ_API_KEY is set (text-only otherwise). Save writes a refs.json manifest next to the images (SKILL CONVENTION) — the source of truth for "what to take from each reference" in multi-reference generation.
serve <path> [--port N] [--no-open] Serve a generated .html (or an output dir) over http://127.0.0.1 and open it. Use this to view previews — file:// URLs are unique origins and break web-font loads, fetch, and extensions.

Read the full file on GitHub · 301 lines

Files

What ships with it

60 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday Changed · +4 lines ba9288eece92
  2. 12d ago First seen · 297 lines · 186 tokens per session scan A 6de2369faf91

Subscribe to this mod's changes

design-tokens is a skill published in the GitHub repository glebis/humane-agentic-design (28 stars, last pushed yesterday), licensed MIT. It adds 186 tokens to every session and 6,071 once invoked, about $0.0009 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

ui-forge

Generate, implement, or redesign distinctive, premium web and app interfaces through product-derived art direction, visual assets, composition, typography, motion, responsive behavior, and anti-slop generation decisions. Use for frontend pages, product screens, dashboards, landing pages, design systems, UI components…

DNM03/ui-forge · 105 tokens

popular-web-designs

54 real design systems (Stripe, Linear, Vercel) as HTML/CSS.

NousResearch/hermes-agent · 23 tokens

web-prototype

General-purpose desktop web prototype. Single self-contained HTML file built by copying the seed assets/template.html and pasting section layouts from references/layouts.md. Default for any landing / marketing / docs / SaaS page when no more specific skill matches.

nexu-io/open-design · 59 tokens

velar-luxury-real-estate

Use this plugin when the user wants a high-end luxury real-estate / architecture landing page with cinematic scroll choreography: a typewriter preloader that lifts away, a scroll-driven house image that rises from below and scales up while pinning to a dark statement section, a sticky dark stats band with count-up…

nexu-io/open-design · 119 tokens

web-prototype-taste-brutalist

Swiss industrial-print web prototype. Newsprint canvas, monolithic black grotesque, viewport-bleeding numerals, hairline grid dividers, hazard-red accent, ASCII syntax decoration. Distilled from Leonxlnx/taste-skill brutalist-skill (Swiss Industrial Print mode).

nexu-io/open-design · 70 tokens

web-prototype-taste-soft

Apple-tier soft web prototype. Silver/cream canvas, double-bezel cards, button-in-button CTAs, generous squircle radii, spring motion, ambient mesh. Distilled from Leonxlnx/taste-skill soft-skill + sections 4–8 of taste-skill.

nexu-io/open-design · 69 tokens