website-qa-sweep template visits the pages you list in an isolated Chromium session, screenshots each one, and reports pages that failed to load, broken links and console errors. It is read only: it clicks nothing that changes anything, and its preset grants no write capability at all.
That makes it the safest Worker to start with, and the one worth pointing at production.
Create and publish
Run it
allowed_domains — a reviewed edit to the draft, published as v2. A page outside the allowlist fails the run with network_destination_denied rather than being fetched quietly.
What you get back
A result object withpages_checked, all_pages_loaded, a per-page array and an issues array; one screenshot artifact per page; and one report artifact summarising the sweep. The run is complete when those validate, not when the model says so.
getArtifact returns metadata and an access decision — never bytes and never a storage location a caller could reach on its own. provenance records who observed the artifact and what the observation is worth: a screenshot of a page is evidence that the page rendered that way for this environment at that moment, and it is not a claim about anything a vendor did.
Try it against Obol’s demo site first
website-qa-sweep supports demo runs. Its sample input sweeps three pages on Obol’s demo storefront, one of which returns a 500 on purpose, so a first run has something real to report.
Watching pages for change instead
competitor-page-watch is the same shape with a different question: it captures the pages you list, compares them against the last capture, and reports what changed, with ignore_selectors for the parts that always move. Also read only, also demo-supported.