What it is.
This is a Claude skill for keeping track of SEO problems on your website. Each time you crawl the site, you give Claude the crawl file, and it lists what's wrong, from broken pages and pages hidden from Google to missing titles.
It also compares each crawl with the last one and marks every problem as new, still there, or gone. A problem only counts as gone when a new crawl confirms it.
Problems on the pages that get the most Google clicks come first, and every serious one needs a person's name on it before the list goes out. Before you start, run one full audit with a checklist like Slate's SEO Audit Template, then use this skill to track what it finds.
What you get.
- A status for every issue against the last crawl: new, still open, resolved, or back again.
- An issue closes only when a recrawl reaches the page and finds the problem gone.
- A coverage warning when a crawl reaches far fewer pages than the last one, so a short crawl can't close issues.
- An accept step for a noindex or canonical that's on purpose, which records who decided and why, then takes the issue off the work list.
- Fourteen audit rules with a written severity for each, from robots.txt blocks down to shared meta descriptions.
- An owner check that fails until every critical and high item has a name on it.
- A worked example over three simulated crawls that runs with no account.
How to set it up.
Install the zip in Claude under Customize > Skills, or unzip it into ~/.claude/skills/ for Claude Code.
Start with one full audit, using a checklist like Slate's SEO Audit Template, so the team agrees on what counts as a problem.
Export each crawl from your crawler, such as Screaming Frog's Internal tab, with one row per URL.
Export the Pages report from Search Console as a CSV, so problems on pages that earn clicks rank first.
Ask Claude to run the audit, with the date the crawl ran. The script refuses a date that isn't later than the last run, so one crawl can't count twice.
Say which noindex and canonical issues are on purpose, then name an owner for each critical and high item left.
Run it again after the next crawl. Only that run can mark issues resolved.
Use cases
Run the audit every sprint
Your team crawls the site every two weeks. Each run shows what last sprint's tickets resolved and what broke since, with no fresh spreadsheet.
Spot what a release broke
A release goes out on Tuesday. The next crawl lists only what's new or back since the last one, so the regression stands out from issues already on the board.
Close tickets with proof
A developer closes the ticket for a broken page. The issue stays open until the next crawl reaches the page and finds it working.
Best for
SEO leads
One person runs the audit each sprint and hands the work list to the team.
Agencies
Each client gets its own folder, and the report shows what last month's work resolved.
In-house dev teams
Tickets arrive one issue at a time, so the recrawl can close each one on its own.