Create a specialized AI agent to automatically investigate and triage bug reports from Sentry. This harness uses the Claude Agent SDK to gather evidence, identify root causes, and generate actionable engineering reports.
Source episode · 07:48
How I Built a Custom AI Harness with the Claude Agent SDK for Bug Triagewith Claire Vo
From 07:48 to 21:00, Claire demonstrates her Sentry triage harness, then walks through its runs, tasks, tools, artifacts, file layout, and investigation result. Clip range: 07:48 to 21:00.
An investigation brief containing confirmed evidence, suspected causes, verification steps, and recommended next actions for a Sentry issue.
Anthropic AI assistant
AI pair programmer
Anthropic AI assistant
OpenAI's cloud-based AI software engineering agent that can execute code, run tests, and handle complex multi-file tasks autonomously.
Step by step
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
6 steps
Before coding, map out the system's architecture. Plan for a frontend like a Terminal UI using Ink, a core built on the Claude Agent SDK, specific tool adapters for APIs like Sentry and Linear, and a file-based artifact store for saving investigation results.
Use an LLM like Claude or Codex to generate the initial code structure. Provide a specific prompt that details the desired SDK, like the Claude Agent SDK, and the intended workflow. The goal is to get a scaffold for an agentic system, not a simple, deterministic script.
Help me build a harness. I want to use the Claude Agent SDK to triage Sentry bugs. Here's the workflow I want it to follow.
Instead of granting general API access, build small, focused modules that connect to tools like Sentry, Linear, and GitHub. These adapters should expose only the specific functions needed for bug triage, such as fetching a particular type of evidence. This constraint makes the agent's work faster and more reliable.
Write a highly specific system prompt and encode it directly into the harness. This ensures the agent receives the same context on every run. The prompt should define the agent's role, the task's scope, which artifacts to use as a source of truth, and the exact format for the expected output.
you are working inside the chat purity engineering harness. It's chat, be specific. It's not an open-ended coding system we wanna use. These artifacts as a source of truth. And here's the plan to attack a very specific problem. And what I want you to return is X, Y, and Z.
Build a simple interface for developers to run the harness. A command-line or terminal UI (TUI) is an effective option, which can be built with a library like Ink for Node.js. The interface should allow an operator to start an investigation by providing an input like a Sentry link.
Start an investigation by providing the harness with a Sentry bug report link. After the run completes, check the artifact store for its output. Review the 'investigation brief,' which should separate confirmed evidence from likely causes, describe how to verify the hypothesis, and state a next action.
Turn an idea into a PRD, user stories, and a plan.
After the steps
How to recover when the loop fails and where human judgment helps.
Recover
Keep building
A guide for solo entrepreneurs to manage business operations using AI. Learn how to source manufacturers, build and maintain an e-commerce site, and replace dashboards with voice commands, all without a human team.
Turn a complex, artistic concept into a manufacturable 3D model. This workflow shows how to use AI to generate a design from a photo and then create CAD files for 3D printing, even for previously 'unmanufacturable' items.
Learn to translate fashion sketches into a full suite of marketing visuals. This workflow covers creating a detailed design spec, generating a core product photo with AI, and iterating for different contexts like runway and influencer shots.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.