So as I've set up multiple OpenClaw(previously) & Hermes Agents(presently) for many clients & myself since Feb of this year, I've learned quite a bit. AI Agents are great, but they shouldn't do/handle everything. Many workflows for clients have rigid workflows that involve a lot of deterministic processes, which require not much guesswork. So if you were to give these workflows to AI agents/LLMs, it could result in some drift over time. That's where creating deterministic scripts & having a source of truth for those deterministic scripts to follow are key. Many embarking on their first builds, that come from a non-technical background like myself, will not even anticipate this and leave their OpenClaw/Hermes Agent/LLM to handle it all. But sadly, that will result in leaks & then ultimately complaints from clients when things start breaking all the time. Not because the Agent Harness is broken or because the AI/LLM models intelligence isn't good enough, but because you left it to handle something that is best left to processes that shouldn't. I know it sounds like I'm repeating this over-and-over, but it's vital everyone understands this. Last thing you'd want is your reputation ruined, your clients unhappy, and to be working tirelessly on fixing things(even if they are paying you). Any other thoughts or feedback? Comment below!