I removed the separate .claude and .codex structures from my workspace and consolidated everything into one .agents system with AGENTS.md as the universal entrypoint. Claude, Codex, Grok, Kimi etc same rules, same context, same gates. No more AI-specific silos.

This is interesting, and I need your help. What I've been doing recently is, when I'm brainstorming, I come up with some ideas, come up with a start of a plan, come up with what it should look like, and then all of a sudden Claude Code starts building it. I'd rather it stop and build an ICM structure first, IF one is warranted. So I put an instruction in CLAUDE.md. It watches what I'm brainstorming, asks whether this is worth an ICM structure, and offers to fire the icm-architect skill: > When Jeff describes a project, assignment, or body of work that fits the ICM > procedure, say so right away, don't wait to be asked, and offer to run the > icm-architect skill. The test is low-bar and affirmative: could this benefit from > a walkable production structure? If it might, flag it. Do not talk yourself out > of it because "it's just one deliverable." A single animation, essay, deck, or app > counts. When genuinely torn, flag it and let Jeff decline; under-firing is the > failure mode to avoid. Global. Every session, every project. Three separate hedges telling it not to chicken out. And now I'm not sure it belongs there at all. Is this even the system's job? There's a decent argument that knowing when to climb from chat to skill to folders is the human's call, and that an agent nagging "want me to build a workspace for this?" is exactly the kind of unhelpful helpfulness we're all trying to design out. Maybe the right answer is that I should have caught it, and the lesson is just that I didn't. Three things I'd like people to poke at: 1. Has anyone built a standing routing rule like this, and did it stay useful or did it turn into noise you learned to ignore? 2. My threshold is "anything past a trivial one-liner," which is deliberately low. Is that the right signal, or is size the wrong axis entirely? Handoff risk might matter more than size. Something I do once but three people touch needs a folder more than something I run weekly and nobody else ever sees.

@Rj Palalon

Most of the time, plain old programming is the better tool. 1. The rules are clear If every step can be written as a rule, AI is not needed. Use: Code 2. The answer must be exact Math, totals, taxes, and dates should always come out right. Use: Code 3. The same input needs the same result Some jobs should work the same way every single time. Use: Code 4. The answer is already in the data If the system already knows the answer, just ask the system. Use: Query/Database 5. You are moving data around Sending data from one system to another usually does not need AI. Use: APIs 6. Only a few cases are unusual Let code handle the normal work and send the odd cases to a person. Use: Human review 7. A mistake could be costly If getting it wrong matters a lot, keep the risky step under tight control. Use: Human approval 8. No real thinking is needed If the job is just following steps, ICM probably adds very little. Use: Code 9. Your software already does it If the feature already exists and works, adding AI just adds another layer. Use: Existing software 10. You just want to use AI Sometimes the new tool is more exciting than the right tool. Use: Ask why When would you choose NOT to use ICM or AI at all?

https://github.com/jtampac