[SmailG]

Problem

The repo has "Automatically delete head branches" on, but 52 merged sync branches had piled up. GitHub skips that cleanup when an app token does the merge (community discussion #63409). The sync's auto-merge is enabled with GITHUB_TOKEN, so every one of those PRs shows mergedBy: app/github-actions. The two sync PRs merged by a person (#3, #60) were cleaned up normally.

Auto-delete also never applies to a closed PR, so the branch of each superseded sync PR that "Close stale sync PRs" closes is left behind too.

The 56 leftover branches were deleted by hand on 2026-10-19. 52 of them still pointed at their merged PR heads. The other 4 had no PR and held only stale rebased copies of the original port commit, whose one unique file, the supertool symlink, had been removed from main on purpose (0e061ce).

Fix

A new step, "Delete finished sync branches", runs before upstream is fetched on every scheduled run, including runs with no upstream change. It uses only gh api, so no upstream code runs with its token.

• It deletes the branch of every merged or closed sync/* PR, if the branch still exists and still points at that PR's head.
• It keeps a branch whose name an open PR uses. Sync branch names are derived from the upstream commit, so a run can reuse a name.
• It keeps a branch that moved after its PR closed, with a warning.
• A missing branch (404) is skipped silently; any other read error and any failed delete is a warning.
• After a delete it removes the local refs/remotes/origin/<branch>. Otherwise a later push --force-with-lease of the same name in this run would fail its lease.
• The step is continue-on-error with a 5-minute timeout, so a cleanup failure or hang never blocks the sync. It sets shell: bash, because the default bash -e has no pipefail and would show a failed gh pr list as a green step.

"Close stale sync PRs" is unchanged. Passing it --delete-branch was the first version of this PR. It was dropped because it would also delete a branch name that the current run is reusing.

Verification

• I ran the step's exact script against this repo under bash -eo pipefail, as Actions runs it, with three throwaway branches and their local tracking refs:
  • sync/manual-2026-10-19, at merged chore: sync with cursor/plugins — chore(pstack): bump to 0.15.28, sync guide with create-formal-verification-skill (#724) #60's head ea04fe4, was deleted and its tracking ref removed.
  • sync/docs-pstack-port-workflow-and-boundary-guidance-238-7-commits, at closed chore: sync with cursor/plugins — docs(pstack): port workflow and boundary guidance (#238) #26's head a2c2ace, was deleted and its tracking ref removed.
  • sync/feat-pstack-add-poteto-help-skill-502, pointed at main instead of chore: sync with cursor/plugins — feat(pstack): add /poteto-help skill (#502) #57's head, was kept with the "moved" warning.
  • The other 52 finished PRs' branches no longer exist and were skipped with no output. The step exited 0.
  • I then removed the kept test branch by hand; only main and this PR's branch remain.
• I ran the jq filter on a fixture. A closed PR whose name an open PR reuses is excluded, a non-sync/ head is excluded, and an open sync/a-2 does not shield a closed sync/a: the match is exact.
• I ran the script with a stub gh to reach the error branches. A 404 is silent, an HTTP 502 read gives "could not read", a failed delete gives "could not delete", and the script exits 0.
• actionlint reports only the SC2129 style note that already exists on main. SC2016 is suppressed in the script, because $open and $name there are jq variables.

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one): 1

• sync

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

• Configuration used: Repository: SmailG/claude-cursor-plugins/.coderabbit.yml
• Review profile: CHILL
• Plan: Advanced
• Run ID: c23e1129-dc43-4fca-9723-cc69f7191592

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}