fix(auto-merge): wait for current-head review and pre-merge gates

[devantler]

Problem

Restricting the shared auto-merge workflow to trusted bot authors (actions#545 / draft #546) closes the human-PR bypass, but the workflow still arms eligible bot PRs immediately on opened, synchronize, or ready_for_review. Branch protection waits for CI, but it does not currently require the current-head CodeRabbit/Codex review or CodeRabbit pre-merge summary.

Live proof: trusted KSail bot PR #6058 was armed at 18:10Z, had zero review objects and no supported pre-merge result, then merged automatically at 18:30Z as soon as CodeQL finished. This contradicts the merged portfolio contract in monorepo#2128: bot auto-merge may be armed only after the current-head pentad is green.

Proposed direction

Fail closed at the shared workflow boundary:

• keep the exact trusted-bot author gate from fix(auto-merge): restrict privileged workflow to trusted bots #546;
• do not call gh pr merge --auto until a current-head CodeRabbit or Codex green review and a supported green CodeRabbit pre-merge result are both proven;
• if those surfaces cannot be re-evaluated reliably from a safe event, stop auto-arming in the required workflow and leave arming to the maintenance agent after its live pentad check;
• never treat missing, stale, inconclusive, or disabled review/pre-merge state as green.

The implementation must account for the current service-setting blockers tracked in monorepo#2123 and ksail#6051; weakening or bypassing the gate is not an acceptable workaround.

Acceptance criteria

• A trusted-bot PR with green CI but no current-head green review is not auto-armed.
• A trusted-bot PR with a green review but missing/stale/non-green pre-merge state is not auto-armed.
• Only the exact current head can satisfy both review surfaces.
• A fully green trusted-bot fixture remains eligible.
• Human and unknown authors remain denied by fix(auto-merge): restrict privileged workflow to trusted bots #546's allowlist.
• Tests cover missing, stale, failed, and green combinations.

Rough size

M: shared workflow state machine and tests, plus coordination with the CodeRabbit settings trackers. A conservative no-auto-arm fallback is smaller and safer than duplicating a fragile summary parser in privileged CI.

Related: actions#545, actions#546, monorepo#2123, ksail#6051, monorepo#2128.

[coderabbitai]

🔗 Related PRs

#402 - fix(enable-auto-merge-on-pr): retry-harden the gh API/merge network calls [merged]
#405 - fix(approve-pr,upsert-issue): retry-harden unguarded gh network calls [merged]
#412 - ci: skip the suite on the release-commit push to dodge the self-pin tag race [merged]
#427 - fix: resolve first-party self-references at the same commit [merged]
#467 - docs: standardize opt-in default-off flag convention for shared CI [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!