What problem does it solve? Bulk imports, deduplication, and mass text edits on a large task bank can silently corrupt thousands of fields, and naive AI-driven sweeps have already damaged content. This Skill enforces a strict safety protocol so every mass operation stays reversible, measurable, and human-reviewed. ## Core Features & Use Cases - Mandatory dry-run workflow: Commands default to --dry-run, require explicit --apply flags, before/after counters, and a human-reviewed HTML diff of 15-20 sample changes. - Reversibility guarantees: Snapshots of old values (JSON id-to-text mapping) plus database backups, idempotency checks, and differential invariant validation (task counts, KaTeX $ parity, brace balance). - Hard prohibitions: Blocks model-driven mass rewrites of statement/answer/solution fields, unreviewed applies, and irreversible task deletion without owner approval. - Use Case: Before running a deduplication or import command on the production task bank, activate this Skill to generate counters, a preview diff, a backup path, and an owner sign-off gate before any data is written. ## Quick Start Review my import command for the task bank and make sure it follows the dry-run, backup, and owner-approval protocol before I run it on production.
