Site Audit Triage

Crawler audits create the wrong incentive. They flag hundreds of "issues" and the naive response — work down the list, fix each one — wastes days and sometimes actively harms the site. Some flagged "errors" are correct UX design (an H1 should differ from the title in many cases). Some are symptoms of one underlying template bug. Some are crawl artefacts that mean nothing. The expensive expert skill is triage: what's real, what's a false alarm and why, and which single root cause is inflating the numbers.

This skill encodes that triage. First it runs a staleness check — acting confidently on a stale crawl is the most damaging failure mode, so the skill explicitly asks for site size and last deploy date before analysing, instead of guessing. If only a SEMrush summary export is available and any issue shows a small failing count, the skill asks for the per-URL detail export before recommending fixes — because "2 of 49" tells you nothing about which two.

Then it groups symptoms into root causes. A missing canonical tag generates duplicate-title, duplicate-H1, duplicate-description, and canonical-mismatch flags simultaneously. Fixing the one tag resolves all four "issue" counts. The skill traces flags back to causes and tells you which single template change collapses the largest batch.

Finally, it produces an explicit ignore list with reasons. This is the part most audits never give you. "Multiple H1s" on landing pages, "H1 differs from title" on product pages, "page depth > 3" on long catalogue leaves — all of these get flagged by crawlers and are often correct as-is. The skill names each false positive and explains why it's deliberately not on the fix list, so you can defend the decision to a client or a stakeholder later.