Review gets easier when the agent and the reviewer read the same standard. No more "that's not how we do it here". The standard already says how we do it here.
The scariest command an AI agent can run isn't rm -rf /. Everyone knows that one. It's crontab -r, one key away from crontab -e, which deletes your whole crontab without asking.
My AI agent tried something risky. The harness stopped it. /review sent the work back. Second try shipped.
That's Kontekstboard in one video:
https://t.co/dQV61l5Opv
#ClaudeCode#buildinpublic#contextengineering
@kidtsang Exactly. AI doesn’t remove the need for shared context, it makes it more important. When the context lives only in peoples heads, both humans and AI are forced to guess
I've spent 10+ years as an enterprise architect. AI coding agents keep teaching me the same lesson: the code is rarely the hard part. The hard part is knowing what "good" means in this codebase. That's a context problem, not a model problem.
Kontekstboard, the kanban board for context-driven AI development.
/breakdown writes the card. /doing loads every attached skill. /review self-checks before handoff.
Create your account and start your 14-day free trial today.
https://t.co/oE9GohjzTG
#ClaudeCode