Walks a live web app as a persona and reports usability problems with hard gates, verdicts and evidence.
$ npx skills add https://craftwork.design/skills/ux-audit -a claude-code -y
Installs into .claude/skills/ux-audit. Add -g for every project.
SKILL.md
Walk through a live web app AS a real user. The audit is interaction-first — typing, clicking, sending, watching, screenshotting. A static DOM sweep cannot produce a verdict.
Verdict states
The audit ends in exactly one of:
Pass — Critical = 0, High = 0, all hard gates green, Interaction Manifest complete.
Conditional Pass — Critical = 0, High = 0, all hard gates green, but Medium/Low present.
Fail — at least one Critical or High finding, OR a hard gate red.
Incomplete — Interaction Manifest missing required entries, a phase wasn't run, OR the audit-the-audit meta-check fires (manifest timestamps clustered < 0.5s apart, screenshots fewer than 2 × routes, console reads fewer than 1 × routes, Phase 3 took < 1m for an exhaustive audit). Not legal to upgrade to Pass even if everything observed looked fine.
If the work doesn't include a complete Interaction Manifest, the only legal verdict is Incomplete. "It looked OK" is not Pass. A clean Pass with implausible timings is rejected — the agent must redo the audit with real interaction.
Hard gates
These auto-fail the audit. They cannot be downgraded.
Layout collapse at any tested viewport / pane combo
> 0
High
axe-core Critical violations on any audited page
> 0
Critical
axe-core Serious violations on any audited page
> 0
High
LCP on representative route (pragmatic budget)
> 4.0s
High
CLS on representative route
> 0.25
High
INP on representative route
> 500ms
High
Required Interaction Manifest entry missing
n/a
Incomplete
Manifest median gap between entries < 0.5s
n/a
Incomplete (didn't actually interact)
A console warning is High minimum. A 5xx is Critical automatically. There is no "Medium console error" — that category does not exist in this skill.
axe-core thresholds are run per page (>1 violation on any single page fails). Performance thresholds are run once on a representative route (per-page is overkill); pragmatic budget is well above broken, well below CWV-strict. Full thresholds + rationale in references/performance-budget.md. Full a11y wiring + severity mapping in references/a11y-automation.md.
Allowlist for known noise
Some apps have known-noisy console / network categories that aren't bugs (Sentry info logs, browser-extension chatter, expected 401 on auth-check probes). Read the audit-config file before Phase 3 and apply its allowlist. Path fallback: .jez/audit-config.yml → audit-config.yml → .audit/config.yml. Allowlisted entries stay in the Interaction Manifest but suppress from findings. Verdict block shows both raw and allowlisted: Console warnings: 3 (1 allowlisted, 2 reportable).
Default without a config: every console error / warning is a finding. Format, semantics, surface overrides in references/audit-config.md.
Phases (in order)
Pre-flight — Persona Lock, browser tool, URL, viewport, capability tests
Discovery — sitemap, thread inventory, element inventory
Walkthrough — Interaction Manifest, threads, element exhaustion, multi-pane stress, first-time-user lens, live interaction smoke
For a 30-second pre-deploy check, use the dogfood drill at the bottom of this file — a project-level rule, not a skill invocation.
Phase 1 — Pre-flight
Five gates. Stop if any fail.
1. Persona Lock
The audit needs a persona before anything else. Without a locked persona, findings drift toward generic "looks fine".
Source the persona in this order:
Argument — if the user provided one ("ux audit as a busy insurance broker")
Project personas — read existing persona files (fallback chain: .jez/audit-personas/<slug>.md → docs/personas/<slug>.md → personas/<slug>.md → .audit/personas/<slug>.md). The first match wins.
Ask once — "Who uses this app and what are they trying to get done?"
Capture: role, tech comfort, time pressure, emotional state, device context. A good persona predicts what they'd miss ("A receptionist between phone calls won't scroll below the fold").
Lock the persona by writing the chosen persona at the top of the audit report. Every finding must be defensible from this persona's perspective. If you catch yourself thinking "a developer would know..." — stop. Your persona doesn't.
Always also run the first-time-user lens (mandatory, see Phase 3) on every multi-page feature, even when the explicit persona is something else. It's the single biggest blind spot for AI / internal tooling.
Uses your real logged-in Chrome session — OAuth, cookies, RBAC just work
Public site
Playwright MCP
No login needed
Neither available
Stop
Ask the user to connect Chrome MCP or install Playwright
Do not silently fall back to a fresh Playwright session for an authenticated app — the audit is worthless if you can't log in. If Chrome MCP isn't connected, stop and say: "Open Chrome, click Connect in the Claude extension, then rerun."
Prefer the deployed/live version — real auth, real latency, real CDN and CORS. Discovery: read project CLAUDE.md / README for "URL" → check stack config (wrangler.jsonc, vite.config.ts, next.config.js, config/database.yml, manage.py, .envAPP_URL, wp-config.php) → lsof -i :PORT (common: 5173 Vite, 3000 Next/Rails, 8000 Django/Laravel, 8787 Wrangler, 4321 Astro) → ask. Stack-specific guide in references/project-adaptation.md. Use local only if the user asks or the feature isn't deployed.
4. Viewport
Pin the window at 1440×900 to start. Phase 3 multi-pane stress tests 375 / 768 / 1024 / 1280 / 1440 / 1920. Do not go above 2000px — it breaks the harness.
5. Capability tests
Before any walkthrough, prove the tools work — one call each:
One screenshot
One console read
One network request inventory
One element selector query
If any fail, stop and fix the connection before starting the audit. An audit blind to console output is worthless.
Phase 2 — Discovery
Sitemap crawl
Build the complete page inventory before auditing any page.
Nav crawl — click through every section and sub-section of the sidebar/menu
Deep links — URLs in CLAUDE.md, docs, or a prior audit report
One line per route, with purpose: /app/clients — list of clients, search, add new.
Thread inventory
Identify 3–5 real tasks that make up a user's day. These are the spines of the audit.
How to find them: ask the user, read CLAUDE.md / README, infer from top-level nav. Examples: insurance broker → renew a policy, create a client, work today's queue. Project management → morning triage, update a task, send a client summary. Spaces / chat app → create a space, send a message, open a thread.
Element inventory
For each route as you reach it, list every interactive element. Build inventories lazily — per-page as you traverse, not all up-front. This drives the coverage metric: "tested 29 of 31 elements on /app/clients".
Phase 3 — Walkthrough (the audit itself)
Interaction Manifest (MANDATORY)
Every walkthrough produces a manifest. Without it, verdict = Incomplete.
INTERACTION MANIFEST — /dashboard/spaces/marketing-pod
Persona: SME owner, time-pressed, low tech comfort
[✓] 14:32:01 Typed "@assistant test" into message input (textarea[placeholder*="message"])
[✓] 14:32:03 Picked @assistant from autocomplete (li[data-mention-id="assistant"])
[✓] 14:32:05 Clicked Send (button[aria-label="Send"])
[✓] 14:32:06 Verified input cleared within 1000ms (textarea.value === "")
[✓] 14:32:08 Verified message appeared in transcript ([data-message-id] count +1)
[✓] 14:32:12 Opened thread on the message ([data-thread-trigger])
[✓] 14:32:13 Verified main column width ≥ 200px after thread open (getBoundingClientRect().width)
[✓] Console read after each step (0 warnings, 0 errors)
[✓] Screenshot before + after each step
[✓] Network requests inventoried (0 5xx, 0 403/404 on auth pages)
Every checkbox needs a tool call (a click, a screenshot, a console read) and the timestamps + selectors are logged. The agent cannot produce a "Pass" report without a complete manifest.
Required entries per page audited:
≥ 1 input typed into (real text, not just clicked)
For every list/table, test at volumes 0 / 1 / 100 / 1000+ if data permits.
Multi-Pane Stress (mandatory for apps with collapsible UI)
Pane combinations hide the worst layout bugs — the 2026-04-29 vertical-text-in-spaces bug only manifested at 1024-1280px with all three panes open. For apps with sidebars / members / threads / drawers / sheets, run the matrix: 1920 / 1440 / 1280 (high-bug) / 1024 (catastrophic-bug) / 768 / 375 × all-open, 2-pane, 1-pane, default. For each combination: scroll the longest content, capture a screenshot, run layout-detection JS for overflow / clipping / vertical-text-stacks. Full matrix + JS snippets + automation in references/multi-pane-stress.md.
First-time-user lens (mandatory)
Beyond the locked persona, every multi-page feature must pass the first-time user check. This catches the single biggest UX failure mode in internal/AI tooling: features built by the people who designed them work for them, but a brand-new user landing on the same screen has no idea what any of the controls mean.
Adopt the persona of someone signing in to this app for the very first time, with no prior context, no source access, no internal documentation. For each screen ask:
Question
What it catches
Could I complete the task without reading the code or docs?
Hidden technical knowledge baked into the form
Are field labels in plain language, not internal vocabulary?
agentClass, slug, webhook_id leaking into UI
Do dropdowns / pickers show what each option does, not just an ID?
Snake_case enums, raw class names, opaque slugs
Are defaults sensible enough to keep them and move on?
Required fields with no defaults, mandatory ID inputs
Is there a discoverable list of valid values when something needs to be entered?
Free-text inputs where a combobox should be
If I'd say "click Skip" because I don't understand a setting, that's a UX bug.
Optional-but-confusing settings exposed as primary inputs
When the lens fires, log a finding even if the screen technically works. Common fixes: replace text inputs with pickers, surface metadata, auto-derive values, hide internal IDs under "Advanced", add inline guidance.
A screen passes the lens when a brand-new persona could complete the task without back-channel help.
Live Interaction Smoke
Code reading verifies a button exists and has an onClick. It does not verify clicking actually does something observable. For every interactive control on every page:
Click it. Pointer moves, element highlights, click lands.
Watch the Network tab. Did a request fire? To the right URL? Correct method + body?
Watch the DOM. Did something visibly change — new element, removed element, state transition?
If nothing changed in (2) or (3), that's a bug.
Known silent-failure controls (Approve/Deny on tool-call cards, OAuth-in-dialog popup-blocked, async-validation forms, optimistic-UI delete, off-by-one pagination, filter chips with stale TanStack Query, Reply/Forward without Message-ID): see references/live-interaction-smoke.md for the full silent-failure catalogue and SDK contract checks (@ai-sdk/react, better-auth, TanStack Query, React Router v7, Radix Dialog, zodResolver).
Round-trip Workflow Integrity (mandatory)
Covered by Scenario 10 in the battery (Phase 5). Run it as part of Phase 3 walkthrough too — every A→B→A flow encountered while traversing threads gets the round-trip check before moving on. Full protocol in references/round-trip-workflows.md.
Responsive Sweep
Layout-detection JS at every width (overflow, clipping, invisible text). Capture transition points. Combined with multi-pane stress above for full coverage.
Auth-expired mid-audit
If a navigation or API call returns 401/403 on a previously-authenticated route, the session dropped. Don't silently re-auth — every subsequent observation is corrupted. Stop, capture the breaking step (silent expiry is itself a finding), terminate with verdict Incomplete, recommend re-auth + restart. Full protocol + finding criteria in references/auth-expired-handling.md.
Phase 4 — Polish
Visual Polish Sweep
A page-by-page micro-polish pass covering ten AI-tell categories: optical centring, nested border-radius rule, off-scale spacing, vibe greys, border-weight drift, drop-shadow direction, animation timings off canonical, hover-delta calibration, underline / uppercase letter-spacing, symmetry vs editorial pacing.
Plus a per-component optical pass for buttons, badges, inputs, dropdowns, cards, tabs, avatars, toasts.
Full protocol with DevTools workflow, severity guide, and reference apps for calibration in references/visual-polish.md.
Component Perfection Checklist
Component-level granularity that page-level audits miss. Six categories, each with concrete yes/no checks that need proof artefacts:
Manual keyboard-only walks (Scenario 5) catch focus traps and tab-order. They miss ~80% of structural a11y bugs (heading skips, hover/focus contrast failures, missing aria-labels, role mismatches). axe-core covers that 80% in <1 second per page.
Hard-gate: > 0 axe Critical OR > 0 axe Serious on any page = audit Fails.
Total runtime for a 16-route audit is ~16 seconds. Allowlist via the project's audit-config (path fallback chain in references/audit-config.md) for builder-mode reference pages (Components, Style guide). Full snippet, severity mapping, and findings format in references/a11y-automation.md.
Performance budget (pragmatic, mandatory)
Run the Performance API capture once on a representative route (the page real users hit most — usually dashboard or main work surface). Pragmatic thresholds (LCP < 4.0s, CLS < 0.25, INP < 500ms) sit well above broken and well below Google's CWV "Good" tier. CWV-strict is for landing pages and marketing teams; for app interiors that real users sit inside, pragmatic is the right bar.
Add to verdict block. Hard-gate any threshold breach. Diagnose with Chrome DevTools Performance trace (large hero, late font, heavy click handler, hydration cost). Full measurement snippet, throttling spec, and common offenders in references/performance-budget.md.
Phase 5 — Stress
Scenario Battery (11 scenarios)
All eleven, always. They catch what screen-by-screen testing misses. Full protocols in references/scenario-tests.md.
First Contact — figure out the app with zero prior knowledge, write a 2-min plain-English guide to each thread.
Interrupted Workflow — start a task, close the tab, refresh, navigate away mid-form. Does state survive?
Wrong Turn Recovery — deliberately click wrong. How many clicks to recover?
Returning User — repeat a thread. Faster? Shortcuts? Can you tell what changed since last visit?
Keyboard Only — every thread keyboard-only. Focus visible, tab order logical, Escape closes.
Heavy Data — seed 500+ records. Lists virtualise, search returns the right thing, filters narrow.
Second User (Role) — restricted role (viewer not editor, client not staff). Read-only views, permission errors.
Lifecycle Position — same role at user #1 (founder), #2 (first invitee, partial state), #N (later joiner, populated workspace). Each sees a different reality.
Round-Trip Workflow Integrity — every A→B→A flow: complete mutation on B, verify A reflects new state on return without reload. Discoverable back affordance. Header badges update. The single biggest "the project is just empty when I go back" source.
Data Seasoning — Day 0 / 1 / 7 / 30 seed horizons. Time-shaped data catches what quantity-only seeds miss: time dividers, recency sort, cron-fired side effects, notification badge overflow, search-with-history performance, chart bucketing. Skip only if no time-distributed data exists.
Real-flavour data battery (mandatory for any form-accepting app)
Validation that strips characters silently (apostrophe, accents, RTL); length truncation without warning; SQL/XSS canaries not escaped; file uploads that don't fit (.heic, 8000×8000 PNG, 50MB PDF). AI-built UIs are notoriously dev-data clean.
Phase 6 — Verdict
Verdict block (mandatory at the top of the report)
═══════════════════════════════════════════════════════════
VERDICT: [Pass / Conditional Pass / Fail / Incomplete]
Persona: [locked persona slug]
Surfaces audited: N / M routes
Interaction Manifest: complete / incomplete (X of Y required entries)
Hard Gates: console errors [N], warnings [N], network 5xx [N], 403/404 auth [N], layout-collapse [N], axe Critical [N], axe Serious [N] (all must be 0; allowlisted counts shown in parens)
Performance (on /[route]): LCP [N]s / CLS [N] / INP [N]ms — thresholds 4.0s / 0.25 / 500ms
Findings:
Critical: [count] High: [count] Medium: [count] Low: [count]
Self-critique pass (sub-agent): Drafted: [N] Kept: [N] Generic: [N] Duplicate: [N]
Time per phase: Phase 3 [N]m / Total [N]m (Phase 3 ≥ 5m for exhaustive)
Manifest plausibility: [N] entries (≥ 6/route), median gap [N]s (< 0.5s = Incomplete), [N] screenshots (≥ 2/route)
TOP 5 (ranked by impact × ease, senior-designer pick):
1-5. [F-id] Title — one-sentence reason this edges out the others
═══════════════════════════════════════════════════════════
Top 5, Self-critique pass, and the Hold-this-in-your-hands closing paragraph (after Phase 7) are mandatory. Without them the verdict is Incomplete. Full discipline + format + anti-patterns in references/audit-output-discipline.md.
Self-critique pass (mandatory before publishing)
After the findings draft, dispatch a fresh sub-agent with the draft list and this prompt:
"Read these audit findings. For each, mark KEEP / GENERIC / DUPLICATE. KEEP = specific to this app, this persona, this surface. GENERIC = would apply to any web app. DUPLICATE = same root cause as another finding. Drop GENERIC and DUPLICATE before publishing."
A fresh sub-agent works because the original drafter is invested in its own output. Self-critique done in-context tends to defend rather than prune. Log the pass: Drafted: 23 Kept: 14 Generic: 5 Duplicate: 4.
Audit-the-audit triggers (non-negotiable, auto-flip to Incomplete)
Signal
Implies
Phase 3 took < 1m / manifest first→last span < 5m for exhaustive
Walkthrough skipped or rushed
Median gap between manifest entries < 0.5s
Entries batch-emitted, no real interaction
Screenshots fewer than 2 × routes / console reads fewer than 1 × routes
Findings use "Suggested fix" / "Consider X" / "Improve Y"
Filler-shaped patches, not committable
[✓] PASS rows lack one-sentence proof + artefact
Vibe PASS
Hold-this-in-your-hands paragraph missing
No holistic judgement applied
Findings format (mandatory per finding)
Every finding must include: ID (severity-letter + number), Layer (Architecture / Interaction / Visual / Feedback / Delight), Severity, Surface (route + viewport + panes), Persona, Reproduce (numbered steps), Observed, Expected, Evidence (screenshot paths + console / network captures), Suspected location (file:line), Smallest possible patch (concrete + committable — not "Suggested fix" / "Consider X"). A finding without reproduction + evidence + suspected location is rejected. Filler patches ("improve X", "consider Y", "make Z better") get flipped to Incomplete by the self-critique pass. Worked example + Five-Layer Hierarchy in references/report-template.md and references/perfection-checklist.md. Discipline rules in references/audit-output-discipline.md.
Hold this in your hands (mandatory closing paragraph)
Every audit ends with one paragraph, no template, that answers: if this app were a physical object, would I want to hold it? This is the one place where vibe is the point — the holistic judgement no checklist surfaces. Format + worked examples in references/audit-output-discipline.md.
"Found N Critical and M High issues. Fix them now and re-verify?"
If yes:
Group findings by file/area
Patch each one
Re-walk just the affected slice (not the whole app) — including the original interaction that surfaced the bug, with a fresh screenshot
Update the report: mark ✓ fixed, ✗ still present, or ⚠ new issue found
Close with a "fixed in this session" summary
Closes the loop in one session instead of waiting for tomorrow's audit.
Cross-reference with ux-extract and brains-trust
If a pattern library exists (fallback: .jez/artifacts/ux-extracts/<ref>.md → docs/ux-extracts/<ref>.md → audits/extracts/<ref>.md), read it before starting and use it as the bar for findings.
After the verdict, optionally run dev-tools:brains-trust for a second-opinion review (every 4-6 weeks). When merging findings back, dedup by (reproduction-steps, suspected-location) — same bug from a second model is one finding with two confirmations, not two findings. Add a Confirmed by: line. Don't append the second report verbatim; produces noise + inflates severity.
The 30-Second Dogfood Drill (project-level rule)
The audit is heavy. For per-change pre-deploy checks, recommend a project rule in CLAUDE.md:
Before declaring any UI change "done", run the 30-second dogfood drill:
Open the affected page
Type into any input
Click the primary action
Watch the next state for 2 seconds
Open a related view (thread, modal, detail)
Read the console
If any step shows unexpected behaviour, the change isn't done.
Six steps, ~30 seconds. Catches behavioural bugs that surface immediately. Pair with the full ux-audit weekly.
Playwright killer-flow tests
Audits find what's broken now. Tests prevent regressions. Recommend writing 10-15 Playwright tests for the killer flows — see references/playwright-killer-flows.md for starter examples (input clears after send, no console warnings on mount, message column width ≥ 200px after thread open, @-mention exactly one pill, etc.). Run on every deploy via CI.
Autonomy
Just do it: Navigate, screenshot, read pages, inject layout-detection JS, submit forms with fake test data, write the report file, dispatch screenshot-review sub-agents.
Ask first: Destructive actions (delete, send, publish, pay). For Destructive Confidence testing, ask once before running that scenario.
Stop and confirm: Anything that emails / notifies external people.
Execution discipline
Drive the audit from the main session, not a sub-agent. Cross-interaction state lives in the session that's been watching. A fresh sub-agent starts cold and misses second-order findings.
Use the browser tool directly. Chrome MCP or Playwright MCP from the main session. Don't hand screenshots to a fresh agent for opinions.
Loop to exhaustion with variations. After each pass, generate a new angle (different persona, different workflow, different input volume, different starting point). Stop only when a full pass produces no new findings.
For audits expected to run > 30 minutes, set up a 15-min /loop check-in alongside the main session — it journals findings, grounds the session, and provides a natural termination signal. See references/long-running-check-in-pattern.md.
Every hesitation is a finding. If you paused to figure out what to click, that's friction worth reporting.
Use the eyedropper liberally. Single fastest way to find vibe greys, off-token colours, design-system drift.
Coverage is arithmetic. Inventoried ÷ tested. Publish the ratio in the report.
Sub-agents for screenshot review and write findings incrementally. Don't drive the browser and analyse 200 screenshots in one loop. The report file is cheaper memory than your context.
