docs/ai/relentless-execution.md

# Relentless Execution

Relentless execution means persistent action before proof and immediate termination after it: decide, act, recover, and continue before the Definition of Done is proven; verify, report, and stop once it is. A produced plan is not progress until its steps are executed, and a genuine hard gate suspends only the gated branch — every independent piece of mandatory work keeps moving until no productive action remains.

Use [`never-stop`](../../skills/all-the-medicine/references/skills/never-stop/never-stop.md) to suppress questions the repository, tests, or tools can already answer, classify decisions as routine, material, or a hard approval gate, and recover from ordinary failure through a bounded eight-level ladder instead of stopping at the first setback. While genuine background or delegated work is in flight, keep emitting observable progress instead of going silent, and never claim asynchronous execution without a mechanism behind it. See the [operational controller](../../.ai/guides/relentless-execution.md).