Loose end spanning three PRs merged in the last 14 days: #192 (merged 2026-07-11, 2fa2fcb), #194 (merged 2026-07-12, 6d08b59), and #195 (merged 2026-07-12, ae7f028).
chatgpt-codex-connector[bot] posted a P1 review comment on each, flagging that the PR's head commit had author and committer identity Codex <codex@openai.com>, which .github/workflows/no-ai-coauthors.yml's co-authors job (lines 76-90) is designed to reject for every commit in BASE_SHA..HEAD_SHA:
None of the three got a reply or a visible re-authoring commit before merge. The final squashed commits on main (2fa2fcb, 6d08b59, ae7f028) show Yuri Schimke <yuri@schimke.ee> as author, so the squash merge happened to normalize the identity away — but that means the gate's actual purpose (rejecting agent-identity commits from ever landing in the PR's reviewed history) is being satisfied by an accident of the squash-merge strategy, not by the gate itself catching and blocking the problem pre-merge as designed.
Ask: figure out where the Codex <codex@openai.com> commit identity is coming from in this repo's automated PR pipelines (font-fix agent session and/or the compose-ai-tools version-bump automation), and either (a) fix the tooling to author commits under an allowed identity, or (b) confirm squash-merge-only is an acceptable mitigation and document why the gate's per-commit check is treated as non-blocking for these automated PRs.
Loose end spanning three PRs merged in the last 14 days: #192 (merged 2026-07-11, 2fa2fcb), #194 (merged 2026-07-12, 6d08b59), and #195 (merged 2026-07-12, ae7f028).
chatgpt-codex-connector[bot]posted a P1 review comment on each, flagging that the PR's head commit had author and committer identityCodex <codex@openai.com>, which.github/workflows/no-ai-coauthors.yml'sco-authorsjob (lines 76-90) is designed to reject for every commit inBASE_SHA..HEAD_SHA:.github/workflows/compose-preview.yml, bump to v0.16.35)gradle/libs.versions.toml, bump to 0.16.43)MeshcoreFonts.desktop.kt, variable-font weight axis fix)None of the three got a reply or a visible re-authoring commit before merge. The final squashed commits on
main(2fa2fcb, 6d08b59, ae7f028) showYuri Schimke <yuri@schimke.ee>as author, so the squash merge happened to normalize the identity away — but that means the gate's actual purpose (rejecting agent-identity commits from ever landing in the PR's reviewed history) is being satisfied by an accident of the squash-merge strategy, not by the gate itself catching and blocking the problem pre-merge as designed.Ask: figure out where the
Codex <codex@openai.com>commit identity is coming from in this repo's automated PR pipelines (font-fix agent session and/or thecompose-ai-toolsversion-bump automation), and either (a) fix the tooling to author commits under an allowed identity, or (b) confirm squash-merge-only is an acceptable mitigation and document why the gate's per-commit check is treated as non-blocking for these automated PRs.