Every check CrawlX runs

A crawl reports findings, not opinions: every finding here comes from a check with a stated threshold and an authoritative source behind it. Checks are grouped by what they examine, and each is labelled with the severity it reports and how much human judgement its fix needs.

The three fixability classes

• Auto-fixable. A codemod can place the change and a guardrail can verify it — a missing canonical, an absent viewport tag, an empty alt attribute. These are the findings CrawlX can open as a pull request.
• Needs review. The machine can draft the change, but a human has to approve the wording or the target: rewriting a duplicate title, choosing which of two near-identical pages survives.
• Developer fix. The change is a routing, hosting or infrastructure decision — a redirect chain, a missing security header, a slow origin. No safe automatic edit exists.

Severity is the level the check reports by default. Where it says Varies, the check picks a level from the evidence — a 404 is critical, a 403 that only blocked the crawler is not.