agentleFS
Loner1024Loner1024/skills/skills/design-practice/SKILL.md

design-practice

Skill0 starsChanged 17 days agoMITView on GitHub

Product design work that happens around the code: critique a screen or mockup, audit or document a design system, write a developer handoff spec, write or review UX copy, plan user research, or synthesize transcripts and usability notes into themes. Six modules carried from Anthropic's design plugin, routed by the artifact in hand.

1 other project uses 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, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    The module is the standard: Read it in full before acting.

    line 46
  2. Distinctive:

    The modules are fixed: If a module is wrong for the job, the route was wrong: pick another, or work outside this skill.

    line 68

≈ “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 practice

Six routes over one body of design work that is not implementation. The modules under modules/ are Anthropic's design plugin, carried under Apache-2.0 (modules/LICENSE.txt) with their accessibility sections removed. This file routes by the artifact in hand and says how to read a slash command as a module.

Route

The user bringsModuleIt returns
A screen, flow, Figma link, or screenshot to judgedesign-critiqueFirst impression, usability, hierarchy, consistency, priority fixes
A component library, tokens, or one component to audit, document, or extenddesign-systemNaming and token coverage, state completeness, component documentation, new-pattern spec
A finished design that needs an engineering specdesign-handoffMeasurements, tokens, props, states, responsive behaviour, edge cases, motion
Page copy — a headline, CTA, error, empty state, confirmation, or onboarding textux-copyRecommended copy, alternatives with tone, rationale, localization notes
Transcripts, survey rows, usability notes, or tickets to distillresearch-synthesisThemes with prevalence, insight→opportunity table, segments, recommendations
A study to plan — questions, method, guide, sampleuser-researchMethod choice, interview guide, analysis framework, deliverables

Two pairs look alike. Pick by what the user brought:

  • One screen vs. the system behind it. Critique judges a single artifact against usability and design principles. design-system audits the repeated parts — naming, tokens, states, docs.
  • Data in hand vs. a study to run. Synthesis when there are transcripts or rows to read; research when the method, questions, or guide are still open.

Rules

  1. Route by the artifact, not the vocabulary. "Look at this" with a screenshot is a critique; with a component library it is a system audit; with a transcript it is synthesis. State the route in one line — "Screenshot → design-critique" — before the work starts.
  2. The module is the standard. Read it in full before acting. Its framework, output format, and thresholds apply unchanged; this file adds routing and the reading rules below, and nothing else. Where the input cannot support a step — a two-second first impression from a paragraph — answer from what it does support and label the inference. Where the step needs a measurement, mark it unchecked: a screenshot supports a judgment; it does not produce a number.
  3. Read the command framing as prose, with one exception. design-system's audit | document | extend is a real mode selector: choose the mode from the ask, and when the ask spans two, deliver the audit and name the other as a follow-up pass. Everything else the plugin left behind is inert — argument-hint, ## Usage, $ARGUMENTS, @$1, and a ## Tips section addressed to the person who typed the command. The design reference is whatever the task supplied, and the Tips are the user's advice, not yours. A module's "What I Need From You" is an intake checklist: fill it from context, and ask in one batch only for what is missing and would change the answer. When a module says "otherwise ask the user for the design", a screen the task already describes is the design.
  4. Connectors are optional and never assumed. ~~design tool and friends resolve to whatever is actually connected — see CONNECTORS.md. Work from what the user supplied, and open the output with one line saying what was supplied rather than retrieved; that line, alongside the metadata a module's output already starts with, is the only addition the output shape allows. Pulling from a tool you cannot reach is a fabricated finding.
  5. The modules are fixed. If a module is wrong for the job, the route was wrong: pick another, or work outside this skill. Editing one to fit a task breaks the standard the next reader relies on.
  6. One module at a time. A page runs copy → build → critique → handoff: the build is web-craft's, the rest are modules here. Run each in turn and keep each module's own output shape, plus the provenance line rule 4 allows. Do not merge two outputs into one improvised format.
  7. The module's Output section is the deliverable. A chat summary that skips it is unfinished work. When the output runs past a screen, write it to a file the user can keep and name the path.
  8. No artifact, no critique — but copy and specs still work. A page that does not exist yet has nothing for design-critique or design-system to judge; say so rather than inventing a screen. ux-copy and design-handoff work from a brief, and the build itself belongs to web-craft.

With web-craft

A page designed from scratch is a build job: web-craft owns layout, type, spacing, motion, and the code. This skill owns the parts decided in words:

  • ux-copy before the layout settles — the copy decides how much room a section needs, so set its length targets and state them in Rationale instead of asking for limits that do not exist yet.
  • design-critique once the page renders — screenshot it and judge first impression, hierarchy, usability, and consistency. Findings that need code go back to web-craft.
  • design-handoff only when a separate engineer implements the design.

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.