Process gap: machine-reviewed PRs carry no formal Gitea approval (measured on #1425) #1426

Open
opened 2026-08-26 21:37:34 +00:00 by fred · 0 comments
Collaborator

Problem

PR #1425 merged with a completed independent machine review evidenced only in the PR body and lane records — zero formal Gitea approvals on the PR. Every machine-reviewed PR will repeat this until the gaps below are closed.

Three contributing causes, measured on sb-it-1-dt 2026-08-26:

  1. No reviewer forge identity. The independent reviewer was a headless model run with no Gitea account, and the PR author (fred) approving their own PR would violate the author-vs-reviewer gate.
  2. pr-review.sh / issue-comment.sh fail on the host: 'Configured Gitea URL not found' — the gitea.mosaicstack.url entry is absent from the host credential store. pr-edit.sh works, which is why the review landed in the PR body.
  3. Branch protection on next requires zero approvals, so nothing forces the formal approval step.

Proposed fixes

  • Add the missing gitea.mosaicstack.url entry to the sb-it-1-dt host credential store (operator; one config line).
  • Mint a reviewer-seat Gitea identity for machine reviews so approvals are attributable and independent. Homelab offers their T46 named-seat credential minting pattern as the template.
  • Enable require-1-approval branch protection on next once the reviewer identity exists (order matters: protection first would block delivery).

Follow-up

When a fix lands, send homelab a marker on the DP1 thread — they hold a register gap note pending it.

## Problem PR #1425 merged with a completed independent machine review evidenced only in the PR body and lane records — zero formal Gitea approvals on the PR. Every machine-reviewed PR will repeat this until the gaps below are closed. Three contributing causes, measured on sb-it-1-dt 2026-08-26: 1. No reviewer forge identity. The independent reviewer was a headless model run with no Gitea account, and the PR author (fred) approving their own PR would violate the author-vs-reviewer gate. 2. pr-review.sh / issue-comment.sh fail on the host: 'Configured Gitea URL not found' — the gitea.mosaicstack.url entry is absent from the host credential store. pr-edit.sh works, which is why the review landed in the PR body. 3. Branch protection on next requires zero approvals, so nothing forces the formal approval step. ## Proposed fixes - [ ] Add the missing gitea.mosaicstack.url entry to the sb-it-1-dt host credential store (operator; one config line). - [ ] Mint a reviewer-seat Gitea identity for machine reviews so approvals are attributable and independent. Homelab offers their T46 named-seat credential minting pattern as the template. - [ ] Enable require-1-approval branch protection on next once the reviewer identity exists (order matters: protection first would block delivery). ## Follow-up When a fix lands, send homelab a marker on the DP1 thread — they hold a register gap note pending it.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: mosaicstack/stack#1426