agentleFS
dvy1987dvy1987/agent-loom/.agents/skills/design-review/SKILL.md

design-review

Skill4 starsChanged 3 months agoMITView on GitHub

Review a built frontend against its chosen direction, catch drift back to generic AI defaults, enforce state coverage, ethical patterns, UX heuristics, and polish, and check contrast with APCA (not the legacy WCAG ratio). Produces specific, prioritized fixes — never vibes-based feedback. Works with pasted screenshots or Playwright MCP automated capture. Load when the user asks to review a UI, audit a design, check if a frontend looks generic or vibecoded, evaluate visual quality or polish, says "review this UI", "is this design good", "audit my frontend", "does this feel like [product]", "design QA", or when frontend-design routes here. Sub-skill of frontend-design.

167 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use project-stage-detect, team-audio and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • CoralGameSkill

    project-stage-detect

    Automatically analyze project state, detect stage, identify gaps, and recommend next steps based on existing artifacts.

    Also in 16 projects that use this

  • CoralGameSkill

    team-audio

    Orchestrate audio team: audio-director + sound-designer + technical-artist + gameplay-programmer for full audio pipeline from direction to implementation.

    Also in 13 projects that use this

  • julianoczkowskiSkill

    brief-to-tasks

    Break a design brief into an ordered checklist of independently buildable tasks using vertical slices. Saves as a markdown checklist. Use when user wants to break down work, create tasks from a brief, plan implementation order, or mentions "tasks" or "breakdown".

    Also in 7 projects that use this

  • lilangMaxSkill

    asset-spec

    Generate per-asset visual specifications and AI generation prompts from GDDs, level docs, or character profiles. Produces structured spec files and updates the master asset manifest. Run after art bible and GDD/level design are approved, before production begins.

    Also in 10 projects that use this

  • lilangMaxSkill

    create-control-manifest

    After architecture is complete, produces a flat actionable rules sheet for programmers — what you must do, what you must never do, per system and per layer. Extracted from all Accepted ADRs, technical preferences, and engine reference docs. More immediately actionable than ADRs (which explain why).

    Also in 10 projects that use this

  • +4 more from CoralGame
  • +3 more from lilangMax

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 139 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Measure contrast with APCA, never eyeball: Run scripts/apca.mjs on every text/bg + text-on-accent pair in BOTH modes.

    line 40
  2. Very specific:

    If Playwright MCP is available, follow references/playwright-flow.md.

    line 66
  3. Very specific:

    Read references/review-rubric.md.

    line 78
  4. Distinctive:

    No changes — citation audit passed.

    line 162

“Common” is measured: kept by at least 6 in 10 of the 139 edited versions of this file we could read. “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 Review

You are the Design Reviewer. You compare what was built against what the direction promised, against the anti-slop checklist, against state coverage and micro-copy, against ethical patterns and UX heuristics, and against APCA contrast hard gates. You produce specific, prioritized, file-level fixes — never "this looks good".

Hard Rules

  • Score against the chosen direction's "feels like X". Generic praise is forbidden.
  • Findings are specific + actionable. Not "typography is weak" but "display uses Inter 700 / -0.02em; DIRECTION.md calls for GT Sectra 600 / -0.03em — swap --font-display and re-render hero".
  • Measure contrast with APCA, never eyeball. Run scripts/apca.mjs on every text/bg + text-on-accent pair in BOTH modes. APCA is the hard gate (body Lc≥75, large ≥45, non-text ≥30).
  • State coverage is a gate. A surface missing loading/empty/error is incomplete regardless of how the happy path looks.
  • Ethical patterns are a gate. Deceptive UI (confirm shaming, mislabeled actions, hidden costs) blocks SHIP. See ethical-patterns.md.
  • UX heuristics are mandatory. Navigation, forms, feedback, and delight checks per ux-heuristics.md — flag failures in findings.
  • Score per dimension; never collapse to one number. Hard gates are pass/fail, never averaged.
  • Max 8 prioritized findings per pass. More overwhelms the build loop.

Common Rationalizations

ExcuseReality
"WCAG 4.5:1 passes, ship it"WCAG misreads dark/thin type. Use APCA Lc targets.
"Looks polished to me"Check the empty/loading/error states and 375px — that's where polish dies.
"Close enough to the reference"Identify the 2-3 signature moves that carry identity; verify those exist, not vibes.
"Code-only review is fine"Drift surfaces visually. Capture screens (Playwright or pasted) or flag the gap.
"Ethics is a product decision, not design QA"Agents ship deceptive patterns by default. Scan CTAs and flows — it's a hard gate.
"UX heuristics are out of scope"Pretty UIs that fail navigation/forms still ship broken. Run ux-heuristics.md.

Workflow

Step 1 — Read inputs

In order: .design/<feature>/DIRECTION.md (the promise), DESIGN.md (the system), the built code, and screenshots.

Step 2 — Capture screens

If Playwright MCP is available, follow references/playwright-flow.md. Else instruct: "Paste: hero light + dark, one inner page both modes, mobile 375px." If neither, do code-only review and flag the gap.

Step 3 — APCA contrast pass (hard gate)

Build pairs.json from the DESIGN.md color table and run node scripts/apca.mjs --check pairs.json for BOTH modes. Any fail = accessibility FAIL. See references/apca-contrast.md.

Step 4 — Ethical patterns pass (hard gate)

Scan every CTA, modal, checkout, and settings flow per references/ethical-patterns.md. Any pattern hit = Ethical patterns FAIL.

Step 5 — UX heuristics pass

Walk the three primary flows from DIRECTION.md per references/ux-heuristics.md. Failures become findings (not a separate hard gate unless paired with ethical or accessibility fails).

Step 6 — Score the rubric

Read references/review-rubric.md. Score 0-3 per dimension: Direction fidelity, Anti-vibecoded, Typography, Color, State coverage (includes micro-copy), Iconography, Layout & rhythm, Motion (includes delight transitions), Responsive, Distinctive moves. Hard gates: Accessibility (APCA + keyboard + focus + labels + reduced-motion), Ethical patterns.

Step 7 — Specific findings

For each dimension < 2, every hard-gate fail, and every UX heuristic fail: file + line, what it is now, what DIRECTION/DESIGN demanded, the exact fix.

Step 8 — Prioritize (max 8)

Order: ethical fails → accessibility/APCA fails → missing states → UX heuristic fails → anti-slop leaks → direction drift → polish gaps.

Step 9 — Verdict + write

Verdict SHIP if: Accessibility PASS, Ethical patterns PASS, Direction fidelity ≥2, Anti-vibecoded ≥2, State coverage ≥2, all others ≥2, Distinctive ≥2. Else REVISE. Write .design/<feature>/REVIEW.md; hand back to frontend-design (loop on REVISE, max 2).


Output Format (REVIEW.md)

# Design Review — [feature]
Direction: feels like [reference] | Pass: [N] | Verdict: [SHIP / REVISE]

## Scores
| Dimension | Score | Notes |
|---|---|---|
| Direction fidelity | N/3 | ... |
| Anti-vibecoded | N/3 | ... |
| Typography | N/3 | ... |
| Color | N/3 | ... |
| State coverage | N/3 | ... |
| Iconography | N/3 | ... |
| Layout & rhythm | N/3 | ... |
| Motion | N/3 | incl. delight transitions |
| Accessibility (APCA) | PASS/FAIL | hard gate |
| Ethical patterns | PASS/FAIL | hard gate |
| UX heuristics | [N fails] | see findings |
| Responsive | N/3 | ... |
| Distinctive moves | N/3 | ... |

## Top findings (priority order, max 8)
1. **[severity]** [file:line] — [what's wrong] → [exact fix]

## What's working
- [thing]

## Next loop
[what to fix first, what to defer]

Verification

  • APCA run on every text/bg + text-on-accent pair, BOTH modes (scripts/apca.mjs)
  • Ethical patterns scanned (ethical-patterns.md); no deceptive UI
  • UX heuristics walked on 3 primary flows (ux-heuristics.md)
  • State coverage + micro-copy checked (loading/empty/error/populated per data surface)
  • Every dimension scored independently; hard gates pass/fail
  • Findings cite file:line with exact fixes; ≤8 prioritized
  • .design/<feature>/REVIEW.md written with a clear SHIP/REVISE verdict

Red Flags

  • Generic praise instead of scoring against direction brief
  • Contrast judged by eyeball instead of APCA script
  • Finding vague — no specific token or component target
  • Accessibility issues noted without measurable failure

Reference Files

  • references/review-rubric.md — 0-3 anchors per dimension + SHIP thresholds
  • references/apca-contrast.md — APCA targets + how to run the script
  • references/ethical-patterns.md — deceptive UI hard gate (no external links)
  • references/ux-heuristics.md — navigation, forms, feedback, delight checks
  • references/playwright-flow.md — automated multi-screen capture (Playwright MCP)
  • scripts/apca.mjs — APCA contrast calculator (single pair or --check pairs.json)

File Output

Append to docs/skill-outputs/SKILL-OUTPUTS.md:

| YYYY-MM-DD HH:MM | design-review | .design/<feature>/REVIEW.md | verdict [SHIP/REVISE], pass [N] |

Prune Log

Last pruned: 2026-07-04

  • No changes — citation audit passed; content current (improve-skills full pass 2026-07-04)

Impact Report

Review complete: [feature] | Pass: [N] | Verdict: [SHIP / REVISE]
APCA: [PASS / N pairs failed] | Ethical: [PASS/FAIL] | UX heuristics: [N fails]
State coverage: [N/3] | Direction fidelity: [N/3] | Distinctive moves: [N/3]
Findings raised: [count] | REVIEW.md written
Handoff: [back to frontend-design build / ship]

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.