[jeremi]

The container pilot previously installed Registry Stack 0.27 while the adaptors targeted newer clients. Upgrade the fresh synthetic deployment and all four adaptor dependency graphs to published 0.38.0, with pinned release checksums and OCI digests.

The pilot now uses stock ThunderID, Casework review followed by explicit BREG application, and Evidence with a persistent non-exportable Transit key. Its runtime, Compose project, host ports and volumes are separate from retained 0.27 data. Explicit client bindings include the private CA, OAuth audience/resource/scopes and governed Evidence verification policy.

The bridge validates canonical 0.38 HookEnvelope deliveries and returns the native notification acknowledgement after OpenFn acceptance. Existing jobs retain their normalized input shape; the legacy bridge body mode remains available. Recovery uses current runtime configuration and retains identities, delivery generations and credential associations.

The supervised local signer policy requires a numeric key version without an ACL version allowlist; the released signer pins and verifies version 1, and startup refuses rotated keys. The operator limitation is documented in deployment/README.md.

Validation:

• 124 Node and 28 Python tests; all four adaptor package dry runs.
• Fresh preparation with published Linux tools and eight Evidence fixtures.
• Released native BREG OAuth/metadata and Evidence send/verify, including missing-source unavailability.
• Native source authorization, intake approval refusal, Casework review and explicit BREG application.
• Complete OpenFn registration, exact/conflicting retries, correction submission, separate approval/application, and verified destination effects.
• Retained service restart and exact replay; native replay of three retained dead letters with generation advancement and payload erasure.
• All three live failure/recovery scenarios: exhausted delivery replay, retry of the failed destination step, and queued work retained across restart.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.