[Governance] Reconcile independent-review policy with a realistic reviewer path

[seonghobae]

Problem

OriginWeave's local quality contract currently lists required independent approval as a PR merge gate, while multiple technically complete PRs are accumulating behind a reviewer path that is not presently provisioned.

This is not ordinary review latency. It is a governance-configuration problem that must be resolved rather than repeatedly reported.

Fresh RCA evidence

The gate must be separated into three different questions instead of assuming they are the same thing:

1. GitHub-enforced approval — current branch protection/rulesets may or may not require a counted approving review. This must be refetched from live policy whenever visibility permits; it must not be invented from PR prose.
2. OriginWeave repository governance — docs/quality-gates.md currently says a PR is mergeable only when required independent approval passes, and AGENTS.md says not to bypass independent approval.
3. Organization solo-maintainer policy — the merged ContextualWisdomLab.github.io PR #138 records the 2026-08-04 organization policy that CODEOWNERS/code-owner approval gates are disabled while the organization has a single maintainer because such gates are unsatisfiable. That policy does not by itself remove every possible non-author review requirement, but it proves that an unsatisfiable review gate must not be treated as a permanent waiting state.

Additional operational evidence:

• a live request for the independent-reviewers team on OriginWeave was rejected by GitHub with HTTP 422 because the team is not a collaborator on this repository;
• opencode-agent currently has repository collaborator permission none on OriginWeave, even though it has submitted formal APPROVED reviews in other CWL repositories;
• previously merged OriginWeave PR feat: authenticate verified TCP peers with TLS #7 contains no formal APPROVED review submissions, so a non-author approval was not GitHub-enforced on that historical merge. This is historical evidence only and must not be promoted to current branch-policy truth without a fresh policy check.

Required outcome

Resolve the contradiction through one durable, empirically valid path. Do not keep technically clean PRs in an indefinite reviewer-provisioning loop.

Path A — provision a real independent reviewer

Use this path if current GitHub policy or a deliberate current OriginWeave/CWL governance decision requires non-author approval.

Acceptable implementations include:

• grant independent-reviewers appropriate OriginWeave repository access and verify GitHub accepts a review request;
• assign another organization team containing eligible non-author reviewers with the required repository access;
• provision one or more eligible non-author collaborators through the normal organization governance process;
• use an existing review App/account only if live GitHub permissions and repository policy prove its formal APPROVED review is eligible and independent.

Path B — explicitly align local OriginWeave governance with the solo-maintainer operating model

Use this path only if live GitHub rules do not require an approving review and the organization intends the 2026-08-04 solo-maintainer policy to apply to OriginWeave's local extra gate as well.

This must be an explicit reviewed governance change, not an implicit bypass. Update the canonical OriginWeave quality/maintainer/automation contracts so that:

• exact-head CI, security scans, 100% owned coverage/rustdoc, resolved valid review findings, exact live-base/lease checks, and branch protection remain mandatory;
• CodeRabbit/OpenCode/GHAS and other automated review evidence remains required where configured but is not mislabeled as an unavailable human approval;
• no self-approval or synthesized identity is introduced;
• the non-author approval gate is marked ON HOLD while only one maintainer exists rather than silently ignored;
• the gate is automatically/reviewably re-enabled only after at least two eligible maintainers/reviewers actually exist;
• the change is linked to the organization solo-maintainer policy and is tested/documented so future automation cannot recreate the impossible gate by assumption.

Feasibility / rejection rules

The following are not remedies:

• repeatedly requesting the already-proven ineligible independent-reviewers route without an access change;
• posting comments or statuses that say APPROVE;
• treating CodeRabbit/OpenCode verdict text as a counted review without eligibility proof;
• self-approval or an alternate author-controlled identity;
• weakening CI/security/coverage or using admin bypass merely to clear the queue;
• continuing to report waiting for approval without testing Path A or Path B against live policy.

Acceptance evidence

The issue closes only after one path is proven end to end.

For Path A:

1. refetch repository/team/collaborator state and prove the reviewer/team is eligible on OriginWeave;
2. on an unchanged PR head, GitHub accepts the formal review request;
3. an eligible non-author formal APPROVED review is visible and satisfies the governing rule;
4. the next technically clean PR can pass the gate without bypass.

For Path B:

1. current GitHub policy is verified not to require a counted approval;
2. the explicit solo-maintainer governance change is merged through normal current repository rules without weakening technical/security gates;
3. canonical OriginWeave documentation and the hourly loop agree on the same conditional approval policy;
4. a technically clean exact-head PR is merge-tested under that policy;
5. a regression contract prevents automation from reintroducing an impossible unconditional reviewer gate while the repository remains solo-maintained.

Operational rule until resolved

This issue blocks only merges whose governing policy truly requires the unresolved approval path. It does not block source fixes, exact-head verification, documentation repair, other PR/issue work, protected-main acceptance, or buyer-visible development. Hourly automation must continue work-conservingly instead of terminating on this condition.

[seonghobae]

Path B is now proven end to end. PR #32 merged to protected main as e27e0039e877ca6c32004dd639ca10f944de87c6, making the repository-local non-author review gate explicitly ON HOLD (solo-maintainer) only when GitHub itself does not require counted approval and fewer than two eligible non-author reviewers exist. The change preserved exact-head CI/security/100% coverage/rustdoc/resolved-findings/live-base/writer-lease/branch-protection gates and added tests/test_review_governance_contract.py to prevent the unconditional impossible gate from returning. No formal review existed on #32, yet GitHub accepted the protected merge, providing current operational evidence that counted approval was not enforced for that merge. A subsequent technically clean PR, #29, passed exact-head CI, Security Scan, and SAST with no unresolved review threads and then merged normally as e029aa2afd6d12b35aaea8159520ba6f8dcd50d2 under the reconciled policy, again without self-approval, synthesized identity, admin bypass, or weakened technical gates. The hourly OriginWeave loop now uses the same conditional approval semantics. This satisfies the issue's Path B acceptance criteria; future live GitHub rules or newly eligible reviewers automatically reactivate the counted non-author gate under the merged contract.