The Steering Loop: Harness Engineering with OpenCode for the Agent-First Era

AI-assisted software development is increasingly limited not by the model, but by the environment around it. When coding agents fail, the root cause is often missing structure, feedback, and engineering constraints.

This lightning talk introduces **Harness Engineering**, the practice of designing repositories and workflows that help AI agents work more reliably. Using **OpenCode**, we’ll explore practical patterns including [AGENTS.md](http://AGENTS.md), LSP integrations, architectural guardrails, and reusable Skills for encoding team knowledge directly into the development workflow.

The talk will provide a concise, practical blueprint for building **model-agnostic, agent-ready engineering environments** that continuously improve as teams learn what works.

Speakers: Soumyadip Chowdhury

Track: [CLS] - OTS