ops-worktrees

Worktrees accumulate because creation is automated and reclamation is not. In this repo three skills run git worktree add and none has ever run git worktree remove; cleanup_worktree.py and audit-worktrees.sh both classify and explicitly delete nothing, and no cron ran either. The count could only go up: 183 registered, 8 created in a single day.

Two distinct losses come out of that, and they need different fixes.

Sprawl is worktrees nobody reclaims. Stranding is work that was never merged — commits sitting in a worktree that no branch name leads back to. The second is the expensive one, and refusing to delete it does not fix it: the work is not deleted, it is simply never found again.

A dirty tree is not a finding