agentleFS
ldaumeldaume/agentic-engineering-harness/skills/design/design-core/SKILL.md

design-core

Skill3 starsChanged 3 months agoMITView on GitHub

Use whenever a UI or UX task has no clearer owner, or spans several surfaces - framing a design problem, choosing between design options, writing a design decision record, defining motion or animation, scoring design quality, or refreshing UI/UX evidence (Designentscheidung, UX-Konzept, Motion, Animation). Routes flows, forms, product UI, websites, dashboards, slides and diagrams, design systems, research, and reviews to their design-* owner. Not for writing code; frontend-craft builds the screen.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every material recommendation carries one level, because a reader must tell a standard from a taste: verified (a standard or test you can …

    line 49
  2. Very specific:

    Write a design decision record for any choice a future reader would question.

    line 75

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Design Core

Turn a design request into a decision about a person's outcome, backed by evidence the reader can check. Start from who is trying to do what, in which context, what friction exists today, and how you will know a change worked, because a screen designed from "make it nicer" optimizes the wrong thing.

Route first

Load the owner of the artifact, because each carries its own golden path and the core alone stays generic. One primary owner per task; add a second only when the task really has two deliverables.

TaskOwnerOften with
discovery, interviews, usability tests, synthesisdesign-research-
a user flow, journey, states, edge casesdesign-flowsdesign-product-ui
a form, validation, errors, submit, destructive actionsdesign-formsdesign-review
SaaS or app screens, navigation, settings, AI-native UIdesign-product-uidesign-forms
website, landing page, marketing or editorial sitedesign-webwriting-marketing for copy
dashboard, chart, table, metric viewdesign-datavizplain-numbers for figures
slides, deck, diagram, architecture or system mapdesign-slideswriting-core for the words
tokens, components, theming, component specsdesign-systems-
audit or critique of an existing UI, flow, deckdesign-reviewthe artifact's owner

Implementing the result is frontend-craft's job; this family decides what to build and how to verify it. Words on the screen, slide, or page belong to the writing-* skills when installed: load writing-core with the design skill whenever the artifact carries text a person reads, because the design pass checks layout and the writing pass is what catches generated-sounding text and translated terms.

Intake

Infer from the repository, ticket, screenshots, and conversation first; ask only when a missing answer would change the design, because a questionnaire delays work the context could answer. Record an assumption instead and mark it. See references/intake-and-modes.md for the context checklist and the six modes (explore, design, build, review, improve, research refresh).

Evidence discipline

Every material recommendation carries one level, because a reader must tell a standard from a taste: verified (a standard or test you can check), supported (several credible signals), context-dependent (helps here, harms there), hypothesis (untested), aesthetic (a deliberate choice with no claim of superiority). Cite the registry, references/sources.json, as [SRC-...]. Never invent research results, user quotes, metrics, awards, or test outcomes; write "not measured" instead. An award is a jury's subjective score, usability included, not tested usability [SRC-WEB-024].

Quality model

Judge ten dimensions separately and show the profile, not one total, because an average hides the one dimension that sinks the release: utility, usability, clarity, accessibility, interaction quality, visual craft, performance, trust, journey continuity, outcome. Scale and gate rules: references/quality-model.md.

Motion

Motion must do a job - feedback, orientation, continuity, attention - or it goes, and every motion has a reduced-motion variant, because vestibular triggers and distraction are real costs [SRC-WEB-002] [SRC-WEB-006]. Durations, easing, and per-surface rules: references/motion.md.

Decisions

Write a design decision record for any choice a future reader would question; template in references/decision-records.md.

Done when

The owner skill's gates pass, each material recommendation has an evidence level and source or is marked hypothesis, accessibility is checked against WCAG 2.2 AA [SRC-A11Y-001], and what automation could not prove is listed for manual review.

Maintaining the knowledge base

references/sources.json is the registry for every design-* skill: id, URL, access date, findings, limits, evidence level, revisit trigger. references/research-refresh.md is the update procedure. The catalog's scripts/check-knowledge.py fails when a citation has no registry entry. tests/routing-cases.json holds the activation cases, and tests/test_routing_cases.py checks that every skill has one.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.