[devantler]

🤖 Generated by the Agentic Engineer

Why

Platform checks wait for a long offline test family before completing manifest validation. That family takes about six and a half minutes in the recorded baseline, extending feedback for changes that require both.

What

Run the complete offline family alongside manifest checks without removing tests or changing which changes are checked. Deployment rules remain unchanged, and incomplete selected checks still block merging.

Fixes #5075

[devantler]

🤖 Generated by the Agentic Engineer

Local verification at 9d8e4e64ec527b07280b55a5a08065a92896b3d9:

• The new boundary tests failed against the unchanged workflow before implementation. The final source passes the complete root-module Go suite with repository-pinned Go 1.26.9, including 54 executions of the actual result guard and nine unsafe dependency/permission mutations.
• The original bridge Bash controls, ShellCheck, workflow lint, formatting and diff checks pass. Both commits are signed.
• A comparison with base 8db9a3c647fd30adbf2b72ca52d67ffc86de3f47 confirms the entire moved step and its three setup steps are unchanged. The remaining manifest steps, triggers, change-selection job, production job and all other jobs are unchanged. The added job has read-only checkout authority.
• The new job is a sibling of manifest validation, not a dependency of the merge-group-only production job. Selected failure, cancellation, unexpected skip, missing selection and unknown result evidence are rejected by required CI.

The workflow combines cleanly with the queued policy split and the complete-Go integration in read-only merge-tree checks. The current hosted run is 39169088324; its outcome, a substantive current-head review and protected merge acceptance are still pending. This is not a delivered performance claim; the parent retains its multi-run post-merge measurement criteria.

[devantler]

🤖 Generated by the Agentic Engineer

Current head 952cad72dd7d9b3dbb1e0ed6b25dce7ff8c54a2e corrects the first hosted failure. Run 39169088324 failed before path classification because a consumer-wiring ablation assumed the aggregate action was steps[0]; the new explicit result guard moved that action to the next step. The production-dependency validator itself passed. The failed run establishes neither bridge execution nor parallel-job acceptance.

The existing 31-case Bash suite reproduced the failure locally. A further negative control exposed a related blind spot: an unrelated action carrying a similarly named input could mask a missing result on the real aggregate. The test now locates and checks the actual aggregate action by identity. All original cases plus that additional control pass: 32/32 on Bash 5 and macOS Bash 3.2. The complete root Go suite, ShellCheck and diff checks also pass on the corrected source. No family, assertion, production dependency or timeout was removed or weakened.

The earlier Bash pass preceded the final aggregate-order adjustment; it was not sufficient verification of that adjustment. This correction reruns the affected suite on the final source rather than carrying that earlier observation forward. The workflow itself is unchanged from the earlier local Go-boundary proof, and a new hosted run plus substantive current-head review are still required. This remains a draft.

Result evidence is scoped to the actual aggregate actions before checking their inputs. The final source combines cleanly with the integrated guard candidate 2465bcf638d9397ea071995501a28d5930e52777 in a native merge-tree rehearsal; all 32 controls pass on both source trees and both Bash versions. Complete local Go validation passes on both final trees. No queue, production or review gate is bypassed.

[devantler]

🤖 Generated by the Agentic Engineer

@coderabbitai full review

[coderabbitai]

⚠️ Action not completed

Review rate limited.

---

Your included review limit is currently reached under our Fair Usage Limits Policy. This review may still proceed through usage-based billing if eligible. Your next included review will be available in 36 minutes.

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 36 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

• Configuration used: Repository YAML (base), Organization UI (inherited)
• Review profile: ASSERTIVE
• Plan: Advanced
• Run ID: 41055838-6ba1-4de0-81e5-a58dab5f9424

📥 Commits

Reviewing files that changed from the base of the PR and between 8db9a3c and 952cad7.

📒 Files selected for processing: 5

• .github/workflows/ci.yaml
• AGENTS.md
• scripts/tests/test-guard-deploy-waits-for-merge-group-validators.sh
• scripts/validate-eks-ci-role-policy/main_test.go
• scripts/validate-merge-group-heal/flux_bridge_test.go

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[devantler]

🤖 Generated by the Agentic Engineer

@codex review

[devantler]

🤖 Generated by the Agentic Engineer

cr:no-gate@952cad72dd7d9b3dbb1e0ed6b25dce7ff8c54a2e — reason=rate-limit. At 2026-10-18 23:00 UTC I read the authenticated CodeRabbit command refusal, created at 22:59:47 UTC and updated at 22:59:57 UTC. It explicitly states that this included review did not run and the next included review is available in 36 minutes. Treating that minute-precision statement conservatively gives 23:36:57 UTC with a one-minute margin. This establishes the refusal for this PR and request, not a green review or a blanket account-wide outage. The live no-gate guard permits the record; no substantive review or finding was discarded. No paid fallback was activated. Current-head CI is green; the ordered next request is Codex, with the draft retained until a substantive current-head review and all normal gates clear.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-18T23:02:52.433626Z	952cad7	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. 🚀

Reviewed commit: 952cad72dd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[devantler]

🤖 Generated by the Agentic Engineer

Readiness at 952cad72dd7d9b3dbb1e0ed6b25dce7ff8c54a2e:

• Exact-head CI 39171076726 passed, with all 19 jobs terminal and all 37 visible PR contexts clear. The independent complete bridge family actually ran for 5m41s alongside manifest validation (31m16s); its strict selected-result guard and final required aggregation both passed. This exercises the changed scheduling behavior as a contributor, rather than inferring it from YAML alone.
• Local RED/GREEN coverage executes the actual shell boundary in 54 combinations and rejects nine unsafe graph/permission mutations. The complete pinned-Go suite, original bridge controls, ShellCheck, actionlint and all 32 deployment/consumer controls pass on the corrected source. Those 32 controls also pass on Bash 3.2. The first failed hosted fixture-builder run remains a failure; it is not reused as acceptance.
• Codex's substantive clean review names 952cad72dd. A fresh complete read found nine conversation comments, no review objects, no threads, and no comment-form or other unresolved finding. CodeRabbit's explicit rate refusal remains no-gate, not approval.
• All four source commits are GitHub-verified. The short repository-template description passes its write contract. Current-head checks and native merge state are CLEAN, with no live competing local holder. The separate required-gate helper reports every readable gate PASS and active configured code-quality analysis UNVERIFIED; that API limitation is not relabeled as a pass. Native CLEAN is the authoritative required-gate observation.
• Current source combines cleanly with current main and the existing CI candidates in signed private integration rehearsals; affected race and deployment controls pass. No assertion, admission family, production dependency, timeout or trust gate was weakened.

Creation-to-required feedback for this standalone candidate was 48m29s and summed occupied non-skipped runner time 70m57s. These are one pre-merge observation, not the combined improvement, billed cost or the three post-delivery measurements required by #4569. All genuine pre-promotion conditions hold; the next step is ordinary protected merge-queue acceptance, not an admin/auto-merge bypass. Delivery remains outstanding until the source head actually merges.