werkrevisor/docs/prs/2026-10-14-PR#105-codex-gate-test.md

PR#105: The First PR Through the Codex Gate

WARNING: This document describes only the change applied in this PR. It may already be outdated once the next PR is merged. Historic PR documentation is not maintained along with new PRs; treat it as a snapshot.

Related Links

• PR#104: [codex] mode = "gate" on werkrevisor.

Problem

Since PR#104 the branch protection of werkrevisor requires the status ai-review/codex. No PR has passed the gate yet; it is not known whether it holds and releases the merge as intended.

Non-Goals

• No change of the runner or of werkrevisor.

Solution

The PR tests the gate in two steps:

1. One commit adds a defect on purpose. An open Codex line comment must hold the merge with ai-review/codex on pending.
2. The next commit removes the defect, and the thread gets resolved. Then ai-review/codex must be success without a human approval, and the automerge must follow.

The squash merge contains neither the defect nor its removal.

The defect was self.reviews_left < 0 instead of <= 0: a run did one review more than max_reviews_per_run. Result of step 1: Codex found it as a line comment, and ai-review/codex was pending; werkrevisor found it too, as ORANGE.

• Werkbaum: #model.codex.gated is in use.
• Version 0.41.4.

Open Questions

None.

Additional Changes

• A test for max_reviews_per_run; with the defect it fails, the CI was green without it.