[discuss] Launched a local Python guard for stale human approvals

A quiet automation bug: a reviewer approves a file, the file changes later, but approved=true stays green. I packaged the deterministic fix as two readable local Python tools that bind approvals to SHA-256 manifests and stop before downstream side effects when inputs drift. The free runnable sample proves approve → edit → stale: https://github.com/NoxyTree/cadence-starter/releases/tag/free-approval-sample-v1.0.0 Approval Guard is a 10-file / 13-test standalone kit: $0.15 with FIRSTSALE today ($1 base). The full 72-file / 31-test two-tool bundle is $1.05 today ($7 base). No AI account, telemetry, hosted service, or subscription. Listing and exact delivery proof: https://tools.launchllama.co/products/cadence-starter-approval-guard I would value feedback on one thing: is the narrow standalone guard the clearer offer, or is the wider workflow-state engine more useful?