I built guardrails for AI coding agents — same governance principle, new domain

I built guardrails for AI coding agents — same governance principle, new domain Over 25 years of working on complex IT programmes I arrived at a principle I now believe is universal: effective intelligent action — whether by a person, a team, or an AI agent — requires five things. Clear direction. Awareness of context — what happened before, what was decided, what failed. Awareness of resource constraints. Awareness of what your actions will affect downstream. And people who are genuinely engaged and capable of acting. Remove any one and the system degrades. I did not derive this from AI theory. I derived it from watching transitions succeed and fail. At Shell I built a governance framework for IT transitions — quality gates, assurance areas, structured handovers. Shell adopted it as the global standard. It has been used for over 1,000 transitions worldwide. When I started building with agentic coding tools I recognised the same failure modes. So I built a framework for that too. The p