Git worktree lifecycle (plain git)

The ticket work loop over a plain git repo, with no bee installed — the fallback bee automates. One linked worktree per ticket, the branch named for the ticket, tracking set so ahead/behind stays visible. Agents commit or push only on the user's explicit request (work).

Fork

git worktree add ../<repo>-TKT-123 -b TKT-123 from the main tree.
git branch -u main TKT-123 — track the forked branch, so status shows ahead/behind.

Work and review

git status -sb / git diff / git log --oneline review the tree (run in the worktree).
git merge main update from the main tree (concurrent commits land there first).
Update and rerun the tests before reporting the work done.

Land and retire

git add <file> / git rm <file> staging.
git commit -m "TKT-123: comment" commits (explicit request only; 1-line message, comment).
Landing is the user's: git merge TKT-123 in the main tree, or an explicitly requested push.
git worktree remove ../<repo>-TKT-123 once merged; a bullshit tree is rm -rf-ed plus git worktree prune.