@Aelithiaa 100% - server-side branch protection is the real "physically impossible" wall.
The hook is the layer before it: catches the force-push locally, before it ever leaves your machine - so you're not relying on having protected every branch on every repo. You want both, honestly.
A Claude Code agent force-pushed over a private repo without permission β Anthropic's own issue tracker, https://t.co/pb7vedUfHL. Not my customer.
Shipped a push-guard hook this week that blocks exactly that: enforced at the tool boundary, not a CLAUDE.md line the model can ignore.
Two ways in: the $49 one-time kit - installer, 8 hooks, doctor, receipts, the manual, updates. Or hire the pipeline: $1,000/mo, async, human-reviewed before merge, 2 slots.
This thread, the landing page, and the installer came out of the same pipeline. https://t.co/bTFJGx8usH
I let my coding-agent stack package and sell itself. Recording attached.
It's a $49 kit: the enforced guardrail hooks, install doctor, and receipts digest I run daily so Claude Code doesn't need babysitting one command at a time.
Every release runs a fresh-HOME install test before it ships. `doctor` then proves your machine end-to-end - settings parsed, hooks registered, plugins resolved. Won't go green? 14-day refund, no argument. Ceiling claim, on purpose: provably safer. Never "unbypassable".