# Audit Facts Are Automatable Replaceability Judgment Is Not ## Claim In a technical-debt audit, the facts that establish whether a component is a problem — is it maintained, does it have known vulnerabilities, how much does it cost per request — can be gathered by a script, but the decision to remove it depends on business context only a person who understands what the system is for can supply. ## Evidence Matt Dorman's WordCamp US 2026 plugin-audit talk scores every plugin on four axes: abandonment, security (CVEs), performance (autoload weight), and replaceability. The first three are automatable lookups; the fourth — can this plugin actually be swapped for core/theme/a leaner tool without breaking something the business depends on — is explicitly reserved for human judgment. Applied to two real client sites (100 plugins combined), the method cut plugin count 48% with zero lost functionality. ## Scope & conditions - Holds for any audit where "is this a problem" (fact) and "is this safe to remove" (consequence) are genuinely separable questions — plugin stacks, dependency graphs, feature flags, subscription/tool sprawl. - Breaks when the removal decision itself is mechanically derivable from the facts (e.g., a plugin with zero installs and a fatal error on activation needs no judgment call) — the split only matters where business context, not more data, is the limiting factor. - Does not claim the automatable axes are effortless to build, only that they are mechanically checkable in a way replaceability is not. ## Links - [[3. Resources/Permanent Notes/Plugin Count Is Not a Performance Metric Work Volume Is]] — because extends: both notes replace a surface-level shortcut (count, or "is it inactive") with per-item inspection; this note adds the automate/human split for who does the inspecting. - [[3. Resources/MOCs/WordPress Hosting Security Stack MOC]] — because applies-in: a plugin-audit methodology that automates CVE/performance scoring while reserving replaceability for a human is directly usable in site security/performance review workflows. <!-- atomicity: 1 idea ✓ --> - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — applies-in: Joins the cluster on where human judgment stays in AI-assisted work.