backup-restore-verification · git:20260902.28b3431 · 2026-09-02 · sha256 6738373a138e0fa2
backup-restore-verification git:20260902.28b3431A
Immutable. This exact content never changes and is served at /api/v1/blob/6738373a138e0fa2.
No tool to install first
claude plugin marketplace add https://markdownregistry.com/mp/a/art_z6ndnnhim3fchy67/6738373a138e0fa22036f2e78814c3b0e93db51af19ad84c68bcbcf89822e60e.json
Then /plugin install backup-restore-verification@markdownregistry-z6ndnnhim3fchy67 in Claude Code. It installs this exact version, the one graded A here, pinned to the commit it was graded at, and leaves it there.
--- name: backup-restore-verification description: Verify that Linux, application, and homelab backups are current, complete, intact, and recoverable through bounded non-destructive restore drills. Use for backup audits, restore readiness, or recovery proof; do not overwrite live data by default. --- # Backup and restore verification Prove recoverability, not merely job success. Default to read-only inspection plus restores into a fresh temporary destination. Never restore over live data, rotate keys, prune snapshots, change retention, or repair a repository without explicit authorization. ## Build the recovery contract Identify the protected system, backup owner, source paths, exclusions, schedule, repository/backend, encryption and credential dependency, retention policy, monitoring signal, recovery point objective, recovery time objective, and documented restore procedure. Distinguish: - configuration copies or Git history; - filesystem snapshots; - deduplicated backup repositories such as Restic or Borg; - database dumps and application-native exports; - raw application-data copies, which may be only crash-consistent. Do not expose repository credentials, passwords, encryption keys, bucket names, private topology, or recovered personal data. Use `sensitive-info-audit` before sharing evidence. ## Evidence ladder Collect the smallest sufficient proof: 1. Confirm the scheduler and last completed run from its authoritative log or repository metadata. 2. Compare expected sources and exclusions with what the job actually mounts or reads. 3. Run the backup tool's read-only repository integrity and snapshot/listing checks when available. 4. Select a representative recent snapshot and restore a bounded sample into a newly created temporary directory with sufficient space. 5. Verify expected files, hashes where meaningful, readability, permissions, ownership metadata, symlink targets, and application-native validation for the sample. 6. Remove the temporary drill only after recording evidence, and say whether removal is recoverable. Do not retain sensitive restored data as a report artifact. A successful restore command is not enough if the restored data cannot be parsed, opened, or used. For databases, prefer a dump validation or isolated disposable restore using the repository's documented engine/version. Do not attach a recovered database to a live application. Use `homelab-stack-triage` when a backup service or storage dependency is failing. Use `managed-config-drift` to compare restored configuration with its intended managed source. Use `migration` when backup format, repository backend, encryption, or schema changes are in scope. Use `workstation-health-triage` when disks, mounts, memory pressure, or host services may invalidate the backup path. ## Result Report coverage, newest usable recovery point, integrity result, restore sample, validation performed, observed recovery time, gaps against the recovery contract, and any destructive or credential-dependent checks not performed. Classify the result as recoverable, partially proven, failed, or blocked by unavailable evidence.