agentleFS
KyaniteLabsKyaniteLabs/tastecheck/skills/design-system-interview/SKILL.md

design-system-interview

Skill12 starsChanged 11 days agoMITView on GitHub

Design-system interview before new frontend builds or redesigns. Use for vague site/app/landing/dashboard requests, generic-looking UI, direction-setting, aesthetic choices, type/color/density decisions, and DESIGN-SYSTEM.md tokens.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 3 are distinctive, and 5 are 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.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Ask one high-consequence fork at a time.

    line 26
  2. Distinctive:

    Stop when buildable: All nine dimensions are decided, delegated, or blocked.

    line 39
  3. Distinctive:

    Use 4–10 exchanges for a full interview.

    line 42
  4. Very specific:

    Use adaptively; references/interview-contract.generated.md is canonical session/dimension authority.

    line 48
  5. Very specific:

    Write DESIGN-SYSTEM.md from assets/DESIGN-SYSTEM.template.md, then give an approvable build handoff:

    line 67
  6. Very specific:

    Read references/session-protocol.md for readiness/resume rules.

    line 76
  7. Very specific:

    Use approved only when confirmed and buildable.

    line 84
  8. Very specific:

    Stop / handoff: Pause when contradictory or trust-critical direction is unresolved (+1 in contract.json).

    line 129

≈ “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 System Interview

Turn a vague frontend brief into buildable decisions: study evidence, recommend a point of view, and ask only questions that materially change the build.

Start with direction, not questionnaire

Inspect product job, audience, content, marks, constraints, locales, and visual references before asking. Open with:

What I see: <the strongest signal in the brief>
My recommendation: <a specific direction>
Why it fits: <the product consequence>
Choose: <fork A and trade-off> / <fork B and trade-off> / redirect me

Ask one high-consequence fork at a time; combine only when one answer settles several decisions. Reflect each answer in one line for correction without rereading.

Interview loop

  1. Read the room. Separate current evidence from residue. If an existing direction covers five dimensions, confirm it and ask only for gaps.
  2. Propose a fork. Give two brief-compatible outcomes that look or behave materially differently. Recommend one and name its trade-off.
  3. Turn language into consequences. Translate “clean,” “premium,” or “bold” into hierarchy, density, material, type, color, or rhythm; do not debate adjectives.
  4. Record the decision. Mark committed, assumption awaiting confirmation, or blocked by contradiction, with supporting evidence.
  5. Stop when buildable. All nine dimensions are decided, delegated, or blocked; then write artifact and handoff.

Use 4–10 exchanges for a full interview. For urgent internal work, offer a short approval pass. In a one-shot/interrupted run, save the decision snapshot and first resume question; assumptions are not approval.

Decisions

Use adaptively; references/interview-contract.generated.md is canonical session/dimension authority. Examples teach format, never taste.

Canonical IDClose with
referencea real artifact and what it earns
personalitya chosen pole, not a middle
aestheticone concrete phrase predicting hierarchy/material
typebinding evidence, display/body stance, language/measure risk
color_modedominant hue, accent job, and light/dark commitment
density_shapedensity, numeric radius range, elevation
structure_rhythmcomposition, motif, and sectional cadence
signatureone memorable move
imagery_iconographysource/treatment or absence; one icon system

Optionally set motion level.

User receives

Write DESIGN-SYSTEM.md from assets/DESIGN-SYSTEM.template.md, then give an approvable build handoff:

  1. Direction: sentence predicting the interface.
  2. Decisions: completed decision map.
  3. Refusals: three to five defaults this product will not use.
  4. Build contract: semantic tokens, structural rhythm, responsive/accessibility constraints.
  5. Next move: first implementation step, or exact blocker and owner.

Read references/session-protocol.md for readiness/resume rules.

| Dimension / conflict | Evidence | Decision or assumption | Consequence | Confirmation / owner |
| --- | --- | --- | --- | --- |
| Alert hierarchy | ... | committed / assumed / blocked | ... | ... |

Use approved only when confirmed and buildable. Use approval-ready when recommendations and tokens await confirmation; do not hand it to implementation. Use blocked when contradiction or missing authority prevents a safe recommendation. Cite non-obvious choices; each token needs a real build job.

The committed direction in one line: "<source-derived aesthetic phrase> — <chosen
hierarchy>, <color role>, <type stance>, <shape/density>, signature = <specific move>."

Hand hue to color-system, type to web-typography, modes to theming, tokens to implementation skills, and the spec to deslop-ui. Keep one DESIGN-SYSTEM.md source of truth.

Reference files

  • references/interview-bank.md — forks and abstention.
  • references/structure-and-rhythm.md — composition/rhythm.
  • references/tokens.md — token architecture.
  • references/decision-records.md — novel cases.
  • references/session-protocol.md — shortcut, contradiction, resume.
  • references/interview-contract.generated.md — generated authority.

Ready-to-build check

Report rows with direct evidence, reason, remediation, and pass/fail/n/a.

CheckStatusEvidence / provenanceReasonRemediation
Nine required dimensions decided
Existing-direction shortcut or full interview justified
Contradictions and trust-critical rationale resolved
DESIGN-SYSTEM.md and canonical token block complete
Downstream handoff is explicit

Build-ready requires every row passing and artifact approved. approval-ready or blocked is a resumable checkpoint: name unresolved owner and next confirmation; do not start implementation. Deliver one-line direction, link artifact, then state next action.

<!-- contract:v1:start -->

Contract (generated)

Canonical detail: contract.json.

  • Route: A new interface or redesign lacks a committed visual direction (+1 in contract.json); avoid: An existing site has sufficient evidence to improve without a new direction interview (+1 in contract.json)
  • Exclude: Do not begin implementation before direction is resolved or explicitly assumed (+1 in contract.json)
  • Stop / handoff: Pause when contradictory or trust-critical direction is unresolved (+1 in contract.json); receives [none] -> sends [art-direction, color-system, component-states, data-viz, deslop-ui, empty-states, form-ux, humanize-copy, micro-motion, responsive-layout, spacing-system, tasteroll, theming, web-typography]
  • Output: A decision-complete design-system artifact with explicit assumptions and readiness state
  • Evidence: table_with_evidence with status, reason, remediation, evidence, provenance.
<!-- contract:v1:end -->

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.