dsh-nuke-plugin

Everything in dsh is a plugin, yet installing and uninstalling them leaves residue in .dsh/ directories, recycle zones, and system TEMP. Manual cleanup risks deleting the wrong thing; script-based cleanup is irreversible and leaves no audit trail. dsh-nuke-plugin turns plugin removal and residue cleanup into a database-grade transactional system: every action goes through validate, preview, execute, and undo stages; directory removal is an atomic rename into a backup zone rather than physical erasure; and you can reverse anything before commit.

Core capabilities span the full lifecycle of perception, decision, execution, recovery, and operations. The perception layer offers residue scanning, dependency detection, orphan identification, and health checks. The decision layer provides blast-radius sandbox simulation, trend regression, disk-full countdown, and a Bayesian self-learning oracle that predicts transaction success rates, expected recovery, and the most fragile step from historical execution data. The execution layer enforces cross-process exclusive locks and automatic Saga rollback on failure. The recovery layer supports WAL crash replay, hash-chain audit verification, and chaos drills that simulate power loss in a sandbox and validate the recovery path end to end.

Ideal for dsh users managing multiple plugin profiles, security teams that need auditable compliance records, and any operations engineer wary of the deleted-it-and-cannot-get-it-back scenario. The zero-physical-deletion, reversible, fully auditable design philosophy turns cleanup from a gamble into an engineering discipline.