Inspector troubleshooting
Common Inspector sticking points — what Protect catches, what the gate enforces, and what neither does.
Overview
Most Inspector confusion comes from expecting it to fix things automatically. It doesn't — by design. This page covers the recurring questions.
Common issues
"Protect flagged something but didn't fix it."
That's expected — Protect is rule-only and reports pass/fail; it never rewrites your content, code, or markup. Fixing (or explicitly overriding) a finding is a deliberate step your team takes.
"A release shipped even though a check was failing."
Check whether that specific check was actually wired into the CI gate — the gate only enforces what's connected to it. A check that exists but isn't wired in won't block anything.
"Protect passed, but the page still looked broken when someone clicked through it."
Rule checks like Protect never load a real browser — that's exactly the gap Playwright smoke & journeys is meant to close, and it's still building today. Static metadata being correct doesn't guarantee a page renders or behaves correctly.
"I need to ship despite a blocked gate."
Override paths exist, but they're deliberate and logged, not a silent skip — that's by design, so a block never just quietly disappears.
"Can an agent check its own work against Inspector automatically?"
That's the target for the Inspector MCP server, which is still building — today it exposes a subset of Protect's checks, not the full surface, and it can't override a CI gate.
Limits & honest scope
Inspector's rule checks catch what they're written to catch — they aren't a substitute for human review on tone, brand fit, or legal review.