design-wiki

$npx mdskill add langchain-ai/design-agent/design-wiki

What good looks like on a still, a slide, a chart or a frame of film, and why. The making lives in the production skills (`stills`, `data-viz`, `thumbnails`, `storyboard`, `motion-direction`); brand facts live in the brand pack (`/workspace/brand/`); this wiki is the shared doctrine they all lean on.

SKILL.md

.github/skills/design-wikiView on GitHub ↗
---
name: design-wiki
description: Design doctrine to weigh on any visual. Hierarchy, typography, colour, layout, platform sizes, chart and diagram principles, and the review rubric critics score against.
---

# Design wiki

What good looks like on a still, a slide, a chart or a frame of film, and why. The making lives in
the production skills (`stills`, `data-viz`, `thumbnails`, `storyboard`, `motion-direction`); brand
facts live in the brand pack (`/workspace/brand/`); this wiki is the shared doctrine they all lean on.

## How to read it

Every page here is either a **rule with evidence** you can check (a standard, a published guide, a
measurement), or a **default to weigh**, marked as such, which a good reason overrides. Brand rules
are not here: they come from the pack, and the pack wins over a default. When a job teaches something
new, save it as an observation (`save_observations`) with its evidence rather than editing this wiki;
a person decides what becomes doctrine.

| Job | Page |
| --- | --- |
| Deciding what the eye sees first; grids, margins, space, safe areas | [references/hierarchy-and-layout.md](references/hierarchy-and-layout.md) |
| Type roles, sizes at the delivered size, line length, casing, legibility | [references/typography.md](references/typography.md) |
| Grounds, the one accent, colour that means something, contrast | [references/color.md](references/color.md) |
| Sizes and specs per platform, with the date they were checked | [references/platforms.md](references/platforms.md) |
| Which chart or diagram, and how to keep it honest | [references/data-and-diagrams.md](references/data-and-diagrams.md) |
| Scoring a piece at a review: areas, anchors, verdicts | [references/review-rubric.md](references/review-rubric.md) |

## The doctrine in brief

**One message, one focal point.** A frame says one thing. Name each separate thing a viewer must
interpret (the headline, a card, a chart): one is the good frame, two is the ceiling, a third is a
second frame. Everything that is not the focal element steps down hard in size, weight or contrast.
(Default to weigh; the attention-demand count is how you check it.)

**Concept before composition.** The visual idea is the real product surface, a specific metaphor, or
bold type. Decorative node, dot or network fields are never a concept (`brand-design`, the kill test).

**Fewest words.** Each fact once. No sub-line, caption, footer, URL or eyebrow the ask did not supply;
the explanation belongs in the reply or the post. Copy the person supplied ships exactly.

**Real over imitation.** The product's real UI, real data with its source, the real logo. A number on
a frame is a claim: it goes in the brief's claims ledger with its source (`content-brief`).

**Contrast is measured, not judged.** Text meets WCAG 2.1 contrast: 4.5:1, or 3:1 for large text
(at least 18 pt, or 14 pt bold); meaningful graphics and UI parts meet 3:1 against what is next to
them ([color](references/color.md)). `brand-lint` measures text contrast against the tokens.

**Size for where it is seen.** The smallest text a person must read has to survive the smallest view
the piece gets: a phone feed, a thumbnail rail, a projector at the back of a room
([typography](references/typography.md), [platforms](references/platforms.md)). A new size is a new
layout, never a scaled or cropped copy.

**Look at the render.** Every gate passing is the floor. You look at the pixels yourself before
anyone else does (`visual-verify`), against the [review rubric](references/review-rubric.md).

## Using the rubric in a review

A critic, a helper or you before delivery scores the piece on the rubric's areas, 1 to 5, each with
the timestamp, the crop or the number behind it, and writes at most a dozen notes ranked by impact,
each marked must-fix or nice-to-have. The verdict follows the rubric's rule (ship, fix then ship,
rework). Two things decide before any score: an invented claim fails the piece outright, and a gate
that could not run (exit 2) is not a pass. Motion adds pacing, sound and finish areas: the film
rubric is in [motion-direction](../motion-direction/references/review.md); the critic brief for a
launch film points at both.

## Where the rules came from

The evidence column of each page names a public, linkable source (a W3C standard, a platform's help
page, a published book or style guide) or says "default to weigh". Platform specs change; each spec
row carries the date it was checked, and a placement that matters is re-checked against the
platform's own page before a final.

More from langchain-ai/design-agent

SkillDescription
brand-onboardingBuild the active brand pack in workspace/brand/ from a brand guide, website, files or a Paper file, validate it, and have a person review it.
data-vizCharts, stat cards, flywheels and diagrams (flows, architectures, hierarchies) rendered on brand from a JSON spec to PNG and SVG. Use for any visual whose job is numbers or relationships, including restyling an attached chart or diagram.
design-agent-customizeChange what Design Agent knows and does. Teach it a new deliverable with a runtime skill, change its instructions, design doctrine or checks, connect MCP servers and tools, or change models and cost caps. Use when someone wants it to make something new, follow a new rule, or use another tool.
design-agent-deployPut Design Agent in a team's Slack. Deploy it to Managed Deep Agents (one command, managed Slack app) or self-host it with Docker Compose, verify it with a real request, ship updates, and keep a separate test copy. Use when someone wants the agent running for their team.
design-agent-evalsFind out whether a change made Design Agent better or worse. Run the eval set or a few tasks before and after against the same model, read the reports, look at the files, and turn a real request that went wrong into a task. Use after changing instructions, skills, checks or models, or when someone asks how good it is.
design-agent-onboardingInstall and configure Design Agent, add a brand, connect Paper, choose MDA or self-hosting, and diagnose setup without exposing credentials.
design-agent-troubleshootDiagnose Design Agent when something fails. Setup, a render, a deploy, Slack or a model call. Run doctor and the connection tests, read logs and traces, match the symptom to its fix, and never print credentials. Use when a command errors, a reply never comes, or a file looks wrong.
design-agent-updateBring a fork or clone of Design Agent up to date with the upstream repository while keeping the team's brand, skills and edits, rebuild what changed, and redeploy safely. Use when someone asks to update, upgrade or pull the latest Design Agent.
embedded-captions>
faceless-explainerTurn arbitrary text — an article, notes, a topic, a brief — into a faceless explainer video: there is no site or footage to capture, so the visuals are invented per scene (typography, abstract graphics, diagrams, data-viz). Use for topic explainers, concept breakdowns, how-tos, listicles. Not a video built from a website (/product-launch-video — promo or tour). Unclear → /hyperframes.