What problem does it solve? After provisioning, deploying, binding secrets, or taking backups, ops documentation often keeps asserting the old world — README files saying "nothing is deployed" while production is live — causing maintainers to treat live environments as inert and make dangerous mistakes. ## Core Features & Use Cases - Same-session supersession sweep: Runs a git grep across all tracked markdown and wrangler.toml files for stale phrases like "not provisioned", "nothing is deployed", and "placeholder" immediately after any world-changing action. - In-place supersession discipline: Replaces stale claims with strikethrough plus dated corrections and evidence pointers, preserving history instead of rewriting it. - Hold-row and config-comment hygiene: Flips BLOCKED hold rows with proof commands and fixes wrangler.toml comments that call real provisioned IDs placeholders. - Use Case: After deploying five workers and recording a passing preflight, sweep the repo so PROJECT-STATE.md, README.md, and runbooks stop claiming production does not exist. ## Quick Start After any deploy or provisioning action, ask the AI to run the ops-record reconciliation sweep across all tracked markdown and wrangler files and supersede any stale claims in place.