Branch: Integrate

Integrate $target into the current branch

Replaces the former git-branch-merge / git-branch-rebase / git-branch-squash trio. The mechanical flow (fetch, integrate, state checks) lives in the script; you handle conflicts, the squash commit message, tests and the push.

$strategy is required (merge, rebase or squash); $target defaults to main when empty.

Steps

1. Check the current branch name and confirm with the user before proceeding.
2. Do NOT create a worktree; work in the current directory.
3. Run "$HOME"/.claude/library/scripts/git-integrate.sh $strategy $target (bare $strategy when $target is empty). Its exit codes:
   • 0: integration done. For squash the branch's changes are staged as one unit: write a single descriptive commit message (conventional commits) and commit.
   • 2: state error (dirty tree, detached HEAD, wrong branch, fetch failure). Report the script's message and stop; fix only what the user asks you to fix.
   • 3: conflicts, left in place. List them, then resolve with judgement about whose change is right: our branch's intent wins for the work this branch is about; the incoming side (usually main) wins elsewhere, since its conflicting change typically carries a fix this branch predates. Never resolve by discarding our changes wholesale, and never by discarding main's either. Then git commit (merge) or git rebase --continue after each (rebase/squash).
4. Run the project's tests to verify the integration.
5. Only when tests pass, push: git push after a merge; git push --force-with-lease after a rebase or squash. Never plain --force.

Red flags

Never: push with failing tests; force-push without --with-lease; run the script on a dirty tree ("stash it yourself" is the user's call, not yours). Always: confirm the branch first; keep conflict resolutions minimal and intent-preserving.