We went from vibe coding to agentic engineering. Next up: building the harness

We realised that vibe coding gave us quick results, but these were unmaintainable over time. So we swung to the other extreme: stepping through every line the agent delivers. Good quality, but you become the bottleneck.

The next step is building the harness: the constraints, skills, hooks, and conventions that make agent output reliable without you having to approve every step. Think of it as designing the guardrails instead of steering the wheel.

Everybody talks like they have agents running wild, but nobody has a finished harness. Every time an agent does something wrong, you don't re-prompt. You add a rule, a check, a convention. Each fix makes every future session better. The harness compounds.

Tools like Claude already have mechanisms in place out of the box. Agent instructions to scope what the agent should and shouldn't do. Skills to give agents tools and teach them specific tasks. Hooks allow for automated checkpoints that bring you in at the right moments.

You're not aiming for a perfect setup. Start with one agent, give them a skill, and add a hook when something breaks. After agentic engineering comes harness engineering. It's how you move from babysitting agents to directing them.