✅ No new issues found.

Reviewed changes — since the prior pullfrog review at 2a02720, the source went through several byte-identical rebases (f1022af → 3f30dde → 8f4c46e); the only real deltas at the current head c9762cd are a webhook-registration argument-order flip and a test header addition.

• On registration flipped to events-first — NewGitHubWebhook now calls On([]string{"push"}, webhook.Typed(s.handlePush)) (and likewise for pull_request and the lifecycle events) instead of the handler-first form. This matches the framework's actual signature On(eventTypes []string, fn HandlerFunc) and directly adopts the suggestion chronark raised on github_webhook.go:33.
• Idempotency-key path now exercised in tests — sendWebhook sets X-GitHub-Delivery: delivery-test, so the integration tests now drive the deliveryID := event.ID branch that attaches WithIdempotencyKey to the Restate Send.

I confirmed run.go and BUILD.bazel are byte-for-byte identical to the last reviewed source, so the only behavior-relevant change is the registration flip, which I verified compiles correctly against pkg/webhook/receiver.go's On(eventTypes []string, fn HandlerFunc). The new delivery header strengthens coverage of the idempotency-key path; it is set but not asserted, which is fine — DeliveryId propagation isn't the subject of these tests.

The two documented behavior changes from the original review are unchanged, and chronark's open thread on argument order is now satisfied by the code (left for chronark to resolve). The separate "why not return a proper 400 on malformed payloads" question remains an author decision and is untouched by these commits.

  ^{｜ View workflow run ｜ Using Claude Opus ｜ 𝕏}