Lightning sessionsRecording60 min

Build a Claude Code Verification Harness

Claude Code writes code that looks right. In one hour we build the harness engineering pattern that makes it prove the code works before you review it. We build static analysis gates, holdout tests the implementing agent never sees, and a Doer and Tester loop that sends failures back with the reason attached. It carries over to Codex and Cursor with small changes to each tool's hook setup.

Watch the recording

The whole session is free on Maven. Sign up with your email and the recording opens.

Watch the recording on Maven

Short on time? Watch it as a free mini course: the same session in short lessons, with chapters and transcripts.

Where it sits in a Software Factory

A coding agent is only as reliable as the system around it. This session builds Verification, the layer marked below.

Boundary
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.
Execution
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.
VerificationThis session
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

Agent-written code often looks right and is not. Reviewing every line by hand does not scale. Verifiers make the agent prove its work before you see it: static checks, tests it never saw and a loop that sends failures back with the reason.

What you will learn

Add static analysis gates
Reject code automatically before it merges or deploys, when it breaks your engineering rules.
Write holdout tests
Check behaviour with tests the implementing agent cannot see or tune itself against.
Build a Doer and Tester loop
Send failed work back to the agent with the reason until it passes or escalates.
Wire the verification harness
Connect gates, holdout tests and the Doer and Tester loop into one run that passes or fails.

The ideas behind it

  • Static gates

    Linters, type checks and rules that run before any review.

  • Holdout tests

    Tests the implementing agent never sees, so it cannot write code to pass them.

  • Doer and Tester loop

    One agent writes, another checks, and failures go back with the reason attached.

The series, in order

Each session builds on the one before. Start anywhere: every recording stays on Maven.

  3. Build a Claude Code Verification HarnessVerification
    This session
  4. Build an Autonomous Claude Code AgentExecution, Verification, Delivery

Before you come

For
Engineers who review agent pull requests that look right and break later, and the leads who set the bar for merging them.
You need
Claude Code. The pattern carries over to Codex and Cursor.
Price
Free on Maven. You sign up with your email.
Can't make it live?
The recording is on Maven.

Questions

Does this work with Codex and Cursor?

Yes, with small changes to each tool's hook setup.

Will it slow the agent down?

Each loop takes time, and saves the time you would spend reviewing broken code.

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.

More Lightning sessions

Build the next layer, live with Param

  1. You watched Build a Claude Code Verification Harness

  2. Free Live LabSecure AI Agents with GuardrailsAgents that each sign in with their own identity, act only within the user's permissions and refuse an injected instruction.Free seat
  3. WorkshopDesign Production-Grade Agentic AI SystemsAn architecture for one real workflow at your company that you can take to a design review: roles, models, context, tools, approvals, failure plan and budget.$497
  4. BuildcampAgentic AI Buildcamp for Engineers & PMsDesign, build and deploy a production agentic AI system in four weeks$997