coupon-domains.provenance.json is the reviewed evidence manifest behind the Chrome and Firefox coupon-domains.js snapshots. The extension never contacts coupon directories at runtime.

• CouponFollow: the fixed numeric and A-Z pages under https://couponfollow.com/site/browse/{0,a-z}/all.
• CouponSwift: the 50 initial merchant websiteUrl records server-rendered by https://www.couponswift.com/stores. The updater intentionally does not call the site's load-more action or fetch individual merchant pages.
• Maintainer-vetted: the regional storefronts in scripts/update-coupon-domains.mjs.

Each domain records per-source firstSeen, lastSeen, and active evidence. Source metadata includes its URL, evidence tier, and current domain count. Inactive evidence remains in the manifest for review history, but only domains with at least one active source are emitted to the extension snapshots.

Refresh from the bounded network sources:

npm run update:coupon-domains -- --as-of YYYY-MM-DD

Verify the manifest and both generated modules deterministically without any network requests:

npm run update:coupon-domains -- --check

The updater rejects unexpected hosts, routes, content types, response sizes, source counts, public/private suffixes, and snapshot churn. An intentional large source migration requires the explicit --allow-large-churn flag and should explain the resulting diff in its pull request.

--bootstrap-from-snapshot exists only for the initial provenance migration. It refuses to run after the manifest exists.

The monthly coupon-domain-refresh.yml workflow runs the refresh and tests, then opens or updates a draft pull request on automation/coupon-domain-refresh. It never pushes generated data directly to main.