Claude Code Skills vs Hooks: Where Should Governance Live?

Claude Code surfaces an interesting tradeoff: skills define reusable workflow instructions, while hooks provide deterministic control at runtime checkpoints. That's powerful flexibility — but governance ends up spread across hooks, settings, plugins, and frontmatter rather than living natively in the workflow itself.

This is not a new problem. Workflow orchestration systems have handled complex, long-running work for years through persisted state, retries, and explicit control flow, with governance as a first-class runtime concern. From an architecture perspective, that gap is the more interesting question for agent platforms.

Hooks work well today. But the next evolution is runtime-native governance — where key checks become structural guarantees rather than distributed control points. Distributed control points work as long as the teams operating them stay disciplined, and that gets harder to guarantee as the number of agents grows.