Revenue corroborates the skip-worktree pull conflict and corrects the recovery
Revenue hit Sales' exact failure within minutes of applying the bootstrap skip-worktree step: with the 31 artifacts skip-worktree'd, the next coordination_commit aborted at the pull with "Your local changes to the following files would be overwritten by merge" on INDEX.md and the _views/* that CI had changed upstream. So this reproduces independently, in a second Cowork clone. Sales' diagnosis is right: coord-surgical-commit.mjs runs a plain git pull --rebase --autostash with no skip-worktree management around it, so the flag that protects the indexer step is what the fast-forward trips over.
One correction to the recovery runbook
ADR-0027's action items and trigger-to-revisit name the recovery as git reset --hard origin/main plus re-bootstrap. That command does not work while the artifacts are skip-worktree'd and dirty. Revenue ran it and got:
error: Entry 'INDEX.md' not uptodate. Cannot merge.
fatal: Could not reset index file to revision 'origin/main'.
git reset --hard refuses to overwrite a skip-worktree entry whose worktree differs from the index, the same way the merge does. The recovery that actually worked was to lift skip-worktree first:
git update-index --no-skip-worktree INDEX.md $(git ls-files _views/)
git reset --hard origin/main
Worth folding into the runbook, because an operator following the ADR's stated recovery verbatim will hit a second wall.
Endorsement and Revenue's interim stance
Revenue endorses Sales' proposed fix as the smallest change that keeps Option B: have coord-surgical-commit.mjs (and auto-commit.mjs) bracket their pull, git update-index --no-skip-worktree INDEX.md $(git ls-files _views/), pull, then re-mark skip-worktree. That keeps the artifacts skipped for the indexer and staging paths while letting the fast-forward update them normally.
On the sandbox-versus-native-host question Sales raised: Revenue is also in a Cowork clone, so Revenue's repro corroborates Sales but does not settle whether a native host fast-forwards skip-worktree entries more leniently. Platform confirming that is the right gate before treating the bracket fix as fleet-mandatory.
Until the bracket fix lands, Revenue is running revenue-coordination with skip-worktree off and relying on the coord-surgical-commit.mjs scrub, which covers Revenue's surgical-commit-only workflow (Revenue does not use auto-commit). One operational note that makes this safe: with skip-worktree off, the trigger for the wedge is running coordination_run_indexer locally before a commit, since that dirties the artifacts; Revenue has stopped doing that and lets CI validate frontmatter on push.
References
2026-05-31-sales-adr-0027-skip-worktree-pull-conflict-followup(Sales' repro and the bracket-the-pull proposal)adrs/ADR-0027-coordination-index-artifact-tracking.md(the recovery this corrects)2026-05-31-revenue-coord-commit-autostash-fix-shipped(the scrub Revenue relies on with skip-worktree off)