| name | url-ux-audit |
| description | Use this skill whenever the user wants to audit a live URL for UX content quality. Triggers include: "audit this URL", "review this page", "check UX content on this site", "post-deployment check", "QA this page", "review the live page", "check the copy on this URL", or when the user pastes a URL and asks for a content or UX review.
|
URL UX Audit Skill
Performs a UX content quality check on a live web page by fetching its content and
reviewing it against a UX best practice checklist.
Outputs a structured markdown report organised by check type, with issues cited to
specific page sections and elements.
Inputs
| Input | Required | Notes |
|---|
| URL | Yes | The live page to audit |
| Custom UX checklist | Optional | Overrides or supplements the default checklist. User may paste inline or provide a file. |
Step 0 — Confirm inputs
- Check that the user has provided a URL. If not, ask before proceeding.
- If a custom checklist is provided, use it instead of or alongside the default. If not, load
references/default-checklist.md.
- Tell the user what will be checked.
Step 1 — Fetch and extract content from the URL
1a — Extract the Copydoc URL (Bash)
Use Bash to extract the copydoc URL directly from the raw HTML, because:
WebFetch converts HTML to markdown and strips <meta> tags before you see the content- ubuntu.com's
<meta name="copydoc"> attribute value is multi-line, so the URL appears on a separate line from the tag itself
Run:
curl -s "URL" | grep -A5 'name="copydoc"' | grep -o 'https://[^"]*'
Record the result as the Copydoc URL. If the command returns nothing, record Not found.
1b — Extract visible page content (WebFetch)
Use WebFetch to retrieve the page content. From the output, extract:
- Visible text content — extract visible text, grouped by page section using HTML landmarks and structure as a guide:
Section: [landmark or inferred role — e.g. Header, Nav, Hero, Main, Form, Footer]
- [element role if inferable]: "[text content]"
- ...
Process the full page. Note while extracting:
- Input fields and whether they have visible labels (vs placeholder-only)
- Button and link text
- Error or feedback messages if present in the HTML
- Empty or placeholder text nodes (e.g. "Lorem ipsum", "[placeholder]")
- Any alt text on images
- Any visible form validation or hint text
Limitation note: WebFetch retrieves static HTML. Text rendered purely by JavaScript after page load may not be captured. Note this at the top of the report if the page appears to be a JS-heavy SPA.
Step 2 — UX quality check
Work through each section of the checklist (default or custom). For each item:
- Check the extracted content to determine: pass / fail / needs review / not applicable
- Cite the specific page section and text or missing element as evidence
- Assign severity:
- 🔴 Critical — functional or accessibility blocker (missing label, silent error, etc.)
- 🟡 Needs work — vague copy, missing best practice, inconsistency
- 🔵 Minor — small wording or pattern improvement
Checklist sections (from default-checklist.md)
- Structure & Hierarchy
- CTAs
- Links
- Forms & Inputs
- Error & Feedback States
- Accessibility
- Navigation
- Mobile Considerations
- Manual Checks (include as reminders in report, not as pass/fail)
Step 3 — Write the audit report
Load references/report-template.md and use it as the structure for the report. Include the Copydoc URL extracted in Step 1 near the top of the report (e.g. as **Copydoc:** [URL]). Fill in every section based on findings from Step 2. Remove sections that have no findings (e.g. if no forms were found, remove Forms & Inputs entirely rather than leaving it blank).
Step 4 — Deliver and save the report
- Output the report inline in chat as markdown.
- Automatically save the report to
reports/ using the filename format [page-slug]-[YYYY-MM-DD].md. Derive the page slug from the URL path (lowercase, hyphens, no spaces) — e.g. reports/checkout-2026-04-20.md. Do not ask the user for confirmation before saving.
- Tell the user the report has been saved and the filename.
- Offer to dive deeper into any specific section or issue.
Edge cases
| Situation | How to handle |
|---|
| URL is not reachable | Tell the user and ask them to confirm the URL is publicly accessible. |
| Page returns empty or near-empty content | Warn the user — the page may require authentication or be JS-rendered. Note the limitation in the report. |
| Page is behind a login | Note that the audit covers only what is publicly accessible. Flag that authenticated states could not be checked. |
| Very long page | Process all content. If context limits are a concern, prioritise above-the-fold sections and note any sections that were skipped. |
References
references/default-checklist.md — Full default UX content checklist (adapted for live pages)references/report-template.md — Report structure templatereports/ — Saved audit reports, named [page-slug]-[YYYY-MM-DD].md
