Build Your First Software Factory Execution Harness
An agent with direct access to your repository can overwrite your work, collide with other tasks and run commands it should never run. In this hands-on session we build the execution layer: an ephemeral Git worktree for each task and sandbox boundaries around what the agent can touch. You leave with a harness pattern you can run on your own repository.
Watch the recording
The whole session is free on Maven. Sign up with your email and the recording opens.
A coding agent is only as reliable as the system around it. This session builds Boundary and Execution, the layers marked below.
Request
Boundary
Context
Skills
Execution
Verification
Delivery
Pull request
LayerWhat it doesWhat it stops
BoundaryThis session
Sets what the agent may touch: which files, which commands, which network.
What it stops: An agent editing files outside its task, or running a command it should never run.
Context
Routes the architecture, past decisions and team rules a task needs to the agent working on it.
What it stops: Code that compiles and ignores how your codebase works.
Skills
Keeps engineering know-how in reusable skills, written once and used by every agent.
What it stops: The same instructions pasted into every prompt, drifting apart.
ExecutionThis session
Runs each task in its own Git worktree, inside the boundary.
What it stops: Two tasks colliding, or an agent changing the branch you are working on.
Verification
Checks the work with static gates, holdout tests and a Doer and Tester loop that sends failures back with the reason.
What it stops: Code that looks right reaching your review.
Delivery
Opens the pull request only when every check passes, for an engineer to merge.
What it stops: A bad pull request that costs the team its trust in the agent.
Why this matters
An agent working straight in your repository can overwrite your changes, collide with other tasks and run commands it should not. An execution harness gives each task its own isolated copy and clear limits, so several agents can work at once and nothing touches your branch.
What you will learn
Create a worktree for every task
Give each agent task its own Git worktree that is created, used and removed cleanly.
Set sandbox boundaries
Decide which files, commands and network access an agent is allowed during execution.
Build the execution harness
Connect task intake, isolation and boundaries into one repeatable run.
Keep your own branch isolated
Let background agents work without changing your working branch.
The ideas behind it
A worktree per task
Each task gets its own checkout on its own branch, created and removed automatically.
Sandbox limits
What the agent may read, write and run, enforced outside the prompt.
Isolation from your work
Your branch and working directory stay untouched while agents run.
The series, in order
Each session builds on the one before. Start anywhere: every recording stays on Maven.
Engineers who want agents running tasks in the background without touching their own branch.
You need
Git, and a coding agent you already use.
Price
Free on Maven. You sign up with your email.
Can't make it live?
The recording is on Maven.
Questions
Do I need Docker?
Not to start. Git worktrees give isolation on disk; containers add more when you need it.
Can several agents run at once?
Yes. That is the point of a worktree per task.
Is it free?
Yes. It is a free Lightning session on Maven, and signing up gets you the recording.
After this session
The Live Labs build one agent system piece by piece, free, with time for your questions. The buildcamp builds all of it with you over four weeks. Join a free Live Lab or see the buildcamp.