Bun example: synthetic screening with evidence kept ready for review

[SanctionsKit]

I'm Alex from SanctionsKit. We published a small MIT-licensed Bun script for anyone wiring a screening step into an onboarding workflow.

It sends a fictional contact to our synthetic sandbox, fetches the evidence for the returned screening ID, and checks that the result and evidence belong to the same contact. Both potential_match and no_match leave the review decision open. An API error never becomes a no-match result.

Source, setup and fictional input

The Windmill setup uses a sanctionskit resource with an api_key field pointing to a secret variable. The script has no package dependencies. Keep the same event_id and contact when retrying; it makes one attempt per request, refuses redirects, and checks response size and evidence identity.

To try the local tests without an account or API key, use Node.js 24 or later:

git clone https://github.com/SanctionsKit/sanctions-kit-examples.git
cd sanctions-kit-examples/integrations/windmill
npm test

The tests use local HTTP fixtures, and the Bun runtime checks use in-memory responses. We have not run it in a hosted Windmill workspace or against the hosted API. The source includes the resource schema and import instructions for that next step; it is not a published Hub item.

The sandbox quickstart covers account and key setup. The sandbox is free and uses synthetic records. Production screening is paid and needs a separate integration review. Feedback on the resource setup and retry handling would be useful.