scripod.com
What a harness is and how to build one with Claude Agent SDK

Overview

Shownote

Highlights

Transcript

Chapters

Pins

What a harness is and how to build one with Claude Agent SDK

How I AI

3 DAYS AGO
How I AI

How I AI

3 DAYS AGO
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

Everybody is saying, “It’s not the model, it’s the harness,” but almost nobody stops to explain what a harness actually is. So I did. I built one live on the show: a Sentry bug-debugging harness for my company ChatPRD, using the Claude Agent SDK, a custom ...

Highlights

This episode demystifies the concept of an AI 'harness' by walking through the live construction of a custom bug-debugging tool. The host builds a terminal-based system using the Claude Agent SDK to automate the entire workflow of triaging, analyzing, and documenting software bugs, moving beyond general-purpose AI coding tools.
00:00
A harness is code around an AI agent that makes it more effective for specific tasks.
03:19
Build a harness for repetitive workflows
04:34
Automating triage, root-cause analysis, and follow-up actions.
06:01
Custom harness automates the entire bug workflow without re-explaining intent.
09:49
The harness automates investigation runs without modifying files
11:04
Harness uses Claude Sonnet 4.6 and connects to Sentry, Linear, and GitHub.
13:45
Codex built the best agent but used the Claude Agent SDK
15:08
The harness is simple, with a high-level index for the TUI and eight files for specific tasks.
16:51
The harness constrains agents to solve specific problems efficiently.
19:18
Harness identified root causes and a blind spot
21:04
A harness is simply structured code around an AI agent.

Chapters

In this episode, we cover:
What is an AI harness?
00:00
When to build a harness
03:19
Why Claire picked bug triage
04:33
Why not just use Claude Code?
06:00
Demo: The custom harness interface
07:48
Architecture: runs, tasks, tools, and artifacts
11:04
Building it with Codex and Claude
13:44
Code map and file layout
15:08
A look at the code
16:51
The live investigation result
19:18
How to build your own harness
21:01

Transcript

Claire Vo: A harness is some code around an AI agent that makes it more effective. Why we've seen people build these specific use case harnesses is sometimes with a specific job, you just want to micromanage a little bit. You just want to be more prescript...