workshop for teams working with ai coding agents
harness
lab
A full-day workshop on working with AI coding agents so anyone can continue where your team left off.
Not prompt demos or hackathon energy. Humans steer, agents execute, the repo carries context, verification carries trust. Success is when another team, teammate, or agent can continue the work without the authors in the room.
The public workshop blueprint: how the day is structured, what belongs in the repo, and what only exists in the live workshop instance.
principles
three rules that keep work with agents legible, verifiable, and maintainable.
map before motion
Before you send the agent on another round, make the repo navigable. A short `AGENTS.md`, a clear entry point, and an obvious next safe step beat one more prompt.
verify before you move on
Anchor each meaningful step in evidence as early as possible. A smaller verified step is worth more than a faster path that falls apart later.
work so others can continue
The next person or the next agent should not have to guess what happened. The repo should make clear what holds, what is actually verified, and what the next safe step is.
day structure
five phases that hold the day together.
opening and framing
Shift from hackathon energy to continuation discipline. Success is not how much code you ship, but whether the next team can pick it up.
context is king
A live demonstration of how better context changes output quality. Why AGENTS.md, a plan, and explicit boundaries come before implementation.
build phase
Teams enter a real repo. Before lunch: AGENTS.md, a plan, one executable check, and the first verified step.
continuation shift
Another team takes over your repo. They read first, diagnose second, change only after they can explain the state. Will your work hold up without you in the room — for another team, teammate, or agent?
reveal and reflection
What helped continuation and what caused friction. Signals from the day become next-week practice.
The workshop does not optimize for feature count. It optimizes for continuation quality.
what it is
Harness Lab is a practical workshop about using AI coding agents in real team software work. The focus is on a way of working that stays understandable, verifiable, and usable through handoffs.
for participants
Participants get room-specific context only after the event code. The public page stays intentionally general and does not expose the live state of a specific workshop instance.
what stays private
Real dates, rooms, team assignments, and facilitation data belong in the private workshop instance. The public repo and homepage stay safe and portable.
workshop blueprint
The workshop method has its own blueprint in the repo. The dashboard operates the live workshop, while the canonical day design, rules, and edit boundaries stay there.