With code you need an antagonistic QA/Security agent team to gate keep bad code. When writing philosophy articles I can't tell you enough how important it is to set up a critic team that watches as you write, blowing holes in your weaknesses in real time. For this Block human shields article I created a Block agent well versed in all his papers on the topic and had him watch as I wrote keeping up a constant barrage when I slipped up. When it was done I created an agent for the paper itself and had it debate the Block agent. It's like hardening your attack surfaces in IT ops. You need to get attacked to learn where your defenses are weak.