Live, open source · SEO
Web Audit Kit
The site-level checklist as a script that FAILS. Around 30 checks over what a page validator never looks at: machine files, the link graph and the delivery.
Most SEO tooling validates one page at a time. Web Audit Kit validates the site as a system, which is where the silent breakages live: a machine-readable file pointing at pages that no longer exist, a page that is indexable but orphaned, assets served without a single cache header, a description that gets truncated in the result.
Key features
- Machine-readable files: every URL resolves, nothing points at a redirect
- Meta per page: title with a term, description within the cut, canonical, Open Graph, one H1
- Link graph: zero broken links, zero orphaned indexable pages
- CSS and motion: composited properties only, reduced-motion escape, dead-CSS detection
- Delivery with --live: cache headers, compression, security headers, real status codes
What it does
It is a single shell script with no dependencies - no Node, no Python, no package manager. All the site-specific detail lives in a config file, so the same script runs unchanged on any static site. It ships with the checklist it automates, and with the half a machine cannot judge: whether the page should exist at all.
The specifics
| License | MIT |
| Dependencies | None - bash, grep, sed, curl |
| Verified on | 40 pages in production |
What it has actually done
It runs over these pages before they go up, and it is the reason several of them changed this month instead of shipping broken. Around 30 checks, verified across 40 pages in production, no dependencies to install.
And the honest part, because a checker is easy to oversell: its entire value is the days it says no. If you run it on your site and it finds nothing, either your site is clean or we are missing a check — and we would genuinely rather hear which one, which is what the button above is for.
What we offer
Free, MIT, no signup and nothing to install. Clone it, point it at a folder and run it. If it finds something on your site that it should not have, that is a check we are missing - tell us and it goes in.
FAQ
How is this different from Lighthouse or a crawler?
Those grade a page. This checks the site as a system, which is where the silent breakages live: a machine-readable file pointing at pages that no longer exist, an indexable page nothing links to, assets served without a cache header. Different question, different answers.
What do I have to install?
Nothing. One shell script - bash, grep, sed, curl. No Node, no Python, no package manager. Everything site-specific lives in a config file, so the same script runs unchanged on any static site.
Does it fix what it finds?
No. It fails. That is the whole design: a report you do not read changes nothing, an exit code in front of a deploy changes everything. The half that proposes the fix is SEO Ops, and it is still in build.
It found nothing on my site. Is it even working?
Run it with --live as well - the delivery checks only run there. If it is still clean, tell us what it should have caught. A check that never fires on anybody is a check we are missing, and that is the point of the button above.
Last updated: