Start here
Agentic Harness Engineering
Get the whole team to a shared understanding of agentic engineering. Start with the 20-minute presentation for the big picture — then take the full course to turn it into practice.
Why two paths? The presentation gets everyone speaking the same language fast — what an agent is, why the harness (not the model) decides your results. The course turns that shared picture into a repeatable practice: the ratchet (every agent mistake becomes one permanent fix), hooks over instructions, spec-driven development, and cross-agent review. Watch together, then learn solo.
Further reading
The course distills these sources. Read the primaries to go deeper on any lesson.
-
Addy Osmani — Agent Harness Engineering
The clearest single essay on treating the harness as a first-class engineering artifact. The ratchet, AGENTS.md, hooks, HaaS. -
Simon Willison — Agentic Engineering Patterns
Practical multi-chapter guide for Claude Code / Codex: how coding agents work, git workflows, subagents, TDD. -
LangChain — The Anatomy of an Agent Harness
Origin of “Agent = Model + Harness”; the 52.8% → 66.5% Terminal-Bench jump attributed to the harness. -
Anthropic Engineering — Harness design for long-running apps
First-party harness design: every component encodes an assumption about what the model can’t do. -
HumanLayer — Skill issue: harness engineering for coding agents
Dex Horthy’s “skill issue” reframe; a working <60-line AGENTS.md. -
Birgitta Böckeler (Martin Fowler) — Harness engineering
The user-facing discipline: what harness engineering looks like from the developer’s seat.
See RESOURCES.md for the complete bibliography (25+ sources, including the open-vs-closed model landscape and the Spec Kit / OpenSpec / agent-skills trio).