Harness engineering gets a course layer

WalkingLabs' Learn Harness Engineering packages AI coding-agent reliability into a structured course: lectures, hands-on projects, and copy-ready templates for AGENTS.md, feature lists, progress files, verification, and automated loops. Why it matters: The practical frontier is shifting from better prompts to better workbenches. Reliable Codex or Claude Code runs depend on explicit rules, persistent task state, reproducible environments, verification gates, and stop conditions around the model.

Try this: Use the course as a repo audit checklist: add or review agent instructions, a feature/task list, a progress handoff file, one canonical validation command, and a definition of done before asking an agent to run a long task.