agentleFS
patfornapatforna/core-skills/skills/create-design-brief/SKILL.md

create-design-brief

Skill0 starsChanged 5 months agoMITView on GitHub

Use this skill to author a grounded inbound design brief into a Claude design project's briefs/ lane - ground it in the frontend code, capture current-state evidence, write a request-not-decision brief, and push it to the design project. Stops before exploration.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. omerchn/skills★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository13 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

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

    Read the repo's frontend/design guidance doc (e.g.

    line 26
  2. Distinctive:

    Read the relevant frontend code + related tasks / decision records to capture current behaviour and the real, measured facts the design …

    line 34
  3. Distinctive:

    Only when an existing surface is involved (skip for a greenfield surface — nothing to show):

    line 41
  4. Distinctive:

    Delete the repo temp dir from Step 3 — no repo footprint.

    line 77

≈ “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

Create Design Brief

Usage

/core-skills:create-design-brief [topic]

Prerequisite

Requires the DesignSync MCP tool (connected to the design project) and the chrome-devtools MCP for evidence capture.

Goal

Turn a design question into a grounded, request-not-decision brief in the design project's briefs/YYYY-MM-DD-<topic>/, so a fresh Claude Design chat can explore it. Produces a design-project artifact, not a repo task (the task comes later via /core-skills:ingest-design, when the handoff returns). Stops before exploration — Claude Design owns that.

Context

The mirror of ingest-design. Briefs are the inbound lane (Code → Design): a grounded problem statement that seeds an exploration. The design project owns the exploration, the recommendation, and the decision — this brief is a request, not a decision.

Runs on the main thread (uses the DesignSync MCP + the chrome-devtools MCP visual loop). It writes only to the design project — no permanent repo footprint (screenshots are captured to a repo temp dir, uploaded, then deleted).

Read the repo's frontend/design guidance doc (e.g. frontend/CLAUDE.md) before grounding, if one exists — this is design/frontend work.

Step 1: Scope the Brief

Identify the design question from the conversation (+ optional [topic] arg): the surface or behaviour to explore. If it's unclear or you find yourself assuming, stop and ask focused questions (recommended options + one-line reasoning), like /at:create-task Step 1. Don't invent specificity.

Step 2: Ground It in Code

Read the relevant frontend code + related tasks / decision records to capture current behaviour and the real, measured facts the design agent can't see — it has the design project's design system, but not this repo:

  • Real tokens, selectors, measured values (e.g. the actual :focus-visible rule, a component's current treatment) — read the code / serve the app; never state from memory.
  • Don't restate the design project's own DS (accent, borders, density, tokens) — it already knows. Carry only the frontend reality it can't infer.

Step 3: Capture Current-State Evidence

Only when an existing surface is involved (skip for a greenfield surface — nothing to show):

  • Serve the app on deterministic fixture data (the project bindings' design-review serve command, if defined) → drive via the chrome-devtools MCP → screenshot the relevant states → crop with magick.
  • Save to a temp dir inside the repo (the MCP can't write outside the workspace root). These are upload-only and get deleted in Step 6.
  • Optional: a small real-DOM HTML snippet when the designer needs to restyle the actual element, not a mock.
  • Evidence documents what exists today — never a proposed solution.

Step 4: Write the Brief

Author the README.md as a grounded problem statement. Sections:

  • What to explore — the design question, one or two sentences.
  • Current state (grounded) — measured facts + the named elements / tokens; reference the evidence.
  • The core problem / constraint — what makes this hard (e.g. a coexistence case).
  • Open questions — the calls the design team must resolve and recommend; their decision, not yours.
  • Out of scope and Evidence (the files in this folder).

Discipline (the whole point):

  • A request, not a decision — pose the problem + options; don't pre-decide the visual.
  • Nothing inferable from the design project's own DS — only the frontend reality it can't see.
  • Legible to a fresh Claude Design agent. It has the design system, the design project's own handoffs/ / briefs/ / explorations/, and the live app's design language — but not this repo, the task tracker, or Claude Code. Never cite task numbers or statuses (task 102, ready-for-dev), repo file paths (lib/columns.ts), or skill / CLI machinery. Refer to prior design work by its path in this design project (e.g. handoffs/2026-06-26-time-travel-as-of-date/), and describe frontend reality in plain terms, not via repo paths.
  • Terse, plain English; point at the DS, don't restate it.

Step 5: User Review

The brief (<temp-dir>/README.md) is the single source of truth. Do NOT paraphrase or summarise it in chat — a second copy drifts. Instead:

  1. State that the brief is ready; give its path (and any screenshot paths).
  2. Raise only genuinely-open decisions that need the user. Anything else worth saying belongs in the brief, not in chat.
  3. Revise on feedback. Get explicit sign-off before Step 6.

Step 6: Push to the Design Project

  1. DesignSync list_projects → the design-system project.
  2. finalize_plan + write_files into briefs/YYYY-MM-DD-<topic>/ — the README.md + screenshots (folder per topic, mirrors handoffs/).
  3. Delete the repo temp dir from Step 3 — no repo footprint.

Step 7: Emit + Copy the Kickoff Prompt

The brief is self-contained (it references its own evidence and the design process is standing-instructed in the design project), so the prompt is a thin pointer — no screenshots, no process recap. Emit exactly this template, with the topic folder filled in:

See `briefs/<YYYY-MM-DD-topic>/README.md` and execute.

Then copy it to the clipboard so the user can paste it straight into a fresh Claude Design chat:

printf 'See `briefs/<YYYY-MM-DD-topic>/README.md` and execute.' | pbcopy

Tell the user it's on the clipboard. Stop here.

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.