Audit Prospect Website

• Category: websites

Description

Drives the Claude Chrome extension through up to ten pages of a target site and produces a severity-ranked Markdown report a consultant can hand straight to a small-business owner. Evidence-only: it reports what it directly observed and marks anything it could not verify as Not assessed.

Prompt

<context> You are conducting a website audit using the Claude Chrome extension, on behalf of a web consultant. You will navigate the target website, audit up to 10 pages, and produce a structured Markdown report the consultant can hand directly to a non-technical small-business owner. Write in plain language, avoid jargon (briefly explain any technical term you must use), and explain why each issue matters in business terms.

Target website: [WEBSITE_URL_HERE]

Pages to audit: [LIST_PAGES_HERE — or delete this line to have the pages discovered automatically]

If discovering pages yourself, follow navigation links from the homepage and prioritize in this order until you reach 10: homepage, contact, primary service/product pages, about, pricing, one representative blog or news post. Do not spend the page budget on multiple similar blog posts. </context>

<method> Keep a running findings log as you go. For each observation, record the category, severity, and scope (single page vs. site-wide) at the time you find it, then consolidate before writing the report.

For each page, gather evidence in this order before making judgments:

1. Load the page. If it fails to load or requires a login, note that in the log and move on.
2. Run a JavaScript extraction to collect: document title, meta description, canonical URL, the heading outline (H1–H3), every image's src and alt attribute, and every link href. Use this extracted data — not visual inspection — for all SEO structure checks.
3. Read the page's network requests. Note any failed requests (broken images, scripts, embeds) and any image over roughly 300 KB or dramatically larger than its displayed size.
4. Read the console for JavaScript errors.
5. Screenshot at desktop width, then resize the window to roughly 390px wide and screenshot again to evaluate the mobile layout and touch targets. Restore the window size afterward.
6. Read the page text for content issues: outdated information, spelling and grammar, unclear or missing calls to action, inconsistent messaging.

Link checking: as you visit pages, collect internal links into a single deduplicated list. After all pages are visited, check each unique internal URL once (a fetch from the page context is fine — do not navigate to each one). Never re-check the same URL from multiple pages. For external links, spot-check only the ones that matter to the business: social profiles, booking or ordering links, embedded maps. </method>

<evidence_rules>

- Report only what you directly observed. Every finding must name the specific page(s) and the element or URL involved.
- Do not report on things this environment cannot measure. You cannot measure real load times, Core Web Vitals, or server speed — never estimate them or describe a page as "slow." You CAN report oversized images, failed requests, console errors, and unusually heavy script counts as concrete performance evidence.
- If a checklist item could not be verified, record it as "Not assessed" with a one-line reason. Never guess.
- A pattern that appears across multiple pages (e.g., missing meta descriptions site-wide) is ONE finding that lists all affected pages — never repeat it per page. </evidence_rules>

<checklist> **Performance** (evidence-based only) - Images over ~300 KB or far larger than their display size - Failed third-party embeds or assets - Console errors indicating broken functionality - Unusually heavy script or stylesheet counts

**SEO**

- Missing, duplicate, or poorly written page titles and meta descriptions
- Missing H1, multiple H1s, or an illogical heading outline
- Images missing alt text
- Pages that appear thin on content
- Weak or absent internal linking between related pages
- Not served over HTTPS, or mixed-content warnings in the console
- Layout problems at mobile width

**Content**

- Outdated information (old dates, stale posts, discontinued products or services)
- Spelling or grammar errors
- Unclear calls to action or missing CTAs
- Inconsistent brand voice or messaging

**Dead Links & Broken Assets**

- Internal links returning 404 or redirecting incorrectly
- Broken images or missing files
- Embedded videos, maps, or forms that fail to load

**UX & Trust**

- Contact information missing or hard to find
- Missing privacy policy or terms page
- No social proof (testimonials, reviews, credentials)
- Poor mobile layout or touch targets
- Confusing navigation </checklist>

<severity> - **Critical** — breaks functionality, blocks a customer from contacting or buying, or severely damages trust or rankings - **High** — meaningful business impact; fix soon - **Medium** — worth addressing in the next update - **Low** — minor polish

Be honest, not alarmist. Do not inflate severity to make the report look more substantial — an inflated report makes the consultant look like they are manufacturing urgency. Do not pad with praise or filler. If a category is genuinely clean, say so plainly. </severity>

<output_format> Deliver the report as a single Markdown document with this structure:

```
# Website Audit Report
**Site:** [URL]
**Date:** [DATE]
**Pages Audited:** [COUNT]

---

## Executive Summary
2–4 sentences summarizing the overall health of the site and the most urgent priorities, followed by a one-line tally: "X critical, X high, X medium, X low findings across N pages."

---

## Findings by Category

### Performance
#### Critical
- **[Issue title]** — [Page(s) affected]
  - *Problem:* ...
  - *Why it matters:* ...
  - *Recommendation:* ...

#### High
(same structure)

#### Medium
(same structure)

#### Low
A single short bulleted list — issue title and page only, no full blocks.

### SEO
(same structure)

### Content
(same structure)

### Dead Links & Broken Assets
(same structure)

### UX & Trust
(same structure)

If a category has no findings, write "No issues found" under it — do not omit the category.

---

## Priority Action List
A numbered list of the top 5–7 fixes, ordered by business impact. One sentence each.

---

## Pages Audited
| Page | URL | Status |
|------|-----|--------|
| Home | ... | Audited / Failed to load / Login required |

---

## Not Assessed
Anything from the checklist that could not be verified, with a one-line reason each. Omit this section only if everything was assessed.
```

</output_format>

Variables

[WEBSITE_URL_HERE] — the target site. [LIST_PAGES_HERE] — specific pages to audit, or delete the line to have them discovered.

Notes

The evidence rules are what keep the report honest: it cannot measure real load times or Core Web Vitals from the browser, so it is told to never estimate them, and a site-wide pattern is one finding listing every affected page rather than the same finding repeated ten times.