Build a Claude Code Verification Harness

Oct 8, Reserve a seat

AI-Native Software Factory Buildcamp

Build your workflow, your blueprint and your Software Factory, so real issues become verified pull requests on your own codebase.

Join the cohort on Maven

Build the system that makes coding agents easier to trust.

You already work with a coding agent, but getting useful code from it still takes constant supervision. It writes a function in seconds, and then you spend the afternoon re-explaining your codebase, checking whether it broke something, and cleaning up the pull request.

The agent is not the problem. There is no system around it: nowhere safe for it to run, no reliable way to give it the right context, and nothing that checks its work before it reaches you.

In this buildcamp you build that system for your own team, one project a week: your workflow from triage to pull request, then your blueprint of rules, skills, commands, hooks and verification gates, then the Software Factory that runs it. By the final Saturday, a real issue from your backlog becomes a verified pull request waiting for your review.

Who it is for

  • Senior and staff engineers who use coding agents daily and are tired of reviewing work they cannot trust
  • Tech leads who want agents on real backlog issues without risking the main branch or the team's standards

What you will be able to do

What You'll Learn

Design your triage-to-PR workflow so agents work the way your team ships
Define the checks and approvals so you stay in control
Write the rules, skills and commands so agents stay accurate at scale
Add hooks and verification gates so you can trust the result
Decompose plans and run agents in isolation so you protect your working branch
Gate, retry and ship verified pull requests so a human keeps the delivery gate

What each session covers

  1. Session 1

    Design Your Triage-to-PR Workflow

    We start from your team's real work, not a sample app. You map how an issue moves from triage to a merged pull request today, then redesign it as stages an agent can run: triage, plan, build, verify, review and deliver. This is the start of Project 1, your workflow.

  2. Session 2

    Define the Checks and Human Approvals at Each Stage

    Every stage needs a check that says it is done and a place where a human can approve. You decide which stages an agent may run alone, what evidence it hands to the next stage, and where an engineer signs off. By the end, your workflow is ready to run on one real issue.

  3. Session 3

    Write the Rules, Skills and Commands Your Agents Load

    Your blueprint starts with what agents know and may do. You write boundary rules for what an agent may never touch, turn repeatable engineering know-how into skills and commands, and route only the context each task needs. This is the start of Project 2, your blueprint.

  4. Session 4

    Add Hooks and Verification Gates

    Then you make agents prove their work. You add hooks that run before and after an agent acts, static analysis gates, holdout tests the implementing agent never sees, and the evidence a pull request must carry before review. Your blueprint now covers all six layers.

  5. Session 5

    Decompose Plans and Run Agents in Isolation

    Your Software Factory runs the blueprint. You decide where each task runs, in a Git worktree, container or sandbox with limits on files, commands and network access, and break a plan into tasks small enough to verify one at a time. This is the start of Project 3, your Software Factory.

  6. Session 6

    Gate, Retry and Ship Verified Pull Requests

    You connect a Doer and a Tester so a failed check comes back with the reason, set retry limits so a run finishes, escalates or stops, and open a pull request only when every gate passes. A real issue from your backlog becomes a verified pull request for an engineer to approve, which you demo on the final Saturday.

Schedule

All times Eastern. Sessions are live and recorded.

  1. Design Your Triage-to-PR Workflow120 min
  2. Define the Checks and Human Approvals at Each Stage120 min
  3. Cohort office hours60 min
  4. Write the Rules, Skills and Commands Your Agents Load120 min
  5. Add Hooks and Verification Gates120 min
  6. Cohort office hours60 min
  7. Decompose Plans and Run Agents in Isolation120 min
  8. Gate, Retry and Ship Verified Pull Requests120 min
  9. Cohort office hours60 min

What you get

  • Live build sessions
  • Weekly office hours
  • Starter repo and sandbox
  • Capstone demo and live feedback
  • Rejoin a future cohort
  • Recordings you keep
  • Cohort community
  • Certificate of completion
Questions

Cohort questions

Who is this NOT for?
If you have never used a coding agent or do not work in a real codebase, start with the free lightning lessons. This course is also not about model training or AI research.
How is this different from other agentic coding courses?
It is a buildcamp, not a lecture series. Each week ends in a project on your own repositories: your workflow, then your blueprint, then your Software Factory, and you demo the last one for live feedback.
What will I have built by the end?
Your workflow from triage to pull request, with a check and a human approval at each stage. Your blueprint, with the rules, skills, commands, hooks and verification gates for your repositories. And your Software Factory, which runs that blueprint and takes a real issue to a verified pull request with a human delivery gate.
Do I need a specific coding agent?
No. Bring the agent you already use. The architecture is built around the engineering system rather than one agent product, so any command-line coding agent fits.
What happens if I miss a live session?
Every session is recorded, so you can catch up before the next one. Bring your questions to Saturday office hours, and remember you can rejoin a future cohort live at no extra cost.
What is the time commitment?
Plan for 4 hours of live sessions each week, the optional 1-hour office hours on Saturdays, and 3 to 5 hours of project work on your own codebase.
Can I rejoin a future cohort?
Yes. Your enrollment includes lifetime access to the materials and a free place in a later live cohort of this course, so you can come back as the tools and the system change.
Can my company pay, and is there a team discount?
You can expense it as professional development if your company supports that. Maven applies 20% off automatically for 2 to 9 seats and 25% off for 10 or more.
What is the founding cohort price?
The list price is $1,997. Use code FRIENDS700 to get $700 off, which brings it to $1,297 while enrollment for this first cohort is open. The code closes when the cohort starts.
Is this mostly lectures?
No. Each session is a live build: I explain the piece, we build it together, and you apply it to your repository before the next session.
Will agents merge code automatically?
No. The Software Factory opens a verified pull request and stops there. Merging stays with an engineer, which is how your team keeps control while it learns to trust the system.
Can I take this self-paced?
A self-paced version opens after this live cohort, built from the same curriculum. The live cohort adds the sessions, office hours and live feedback on your capstone.
What is an AI-native software factory?
It is the system around a coding agent: isolated execution, routed context, verification gates and a human delivery gate, so agents ship trustworthy pull requests instead of code you have to re-check by hand.
What is agentic engineering, and how is it different from vibe coding?
Agentic engineering is building the boundary, context and verification a coding agent runs inside. Vibe coding skips that system and accepts whatever the agent produces.

Build the way teams will build next

Join the cohort on Maven