double-checking-before-acting

Checks a critical agent action against its exact target, expected result, and stop condition before and after execution. Use when an agent is about to make a critical edit, run a command or migration, use a credential or tool, change a dependency or model, make a public claim, or affect a release. Do not use for low-stakes reversible edits.