Step01
Identify Problem Areas
Run the normal test command and capture both standard output and errors to a dated log. Record the test count, failures, and total log lines so later work has a baseline.
IntermediateEngineering
How to Systematically Reduce Technical Debt Using AI Agents
Capture a noisy test run, have Claude group warnings by type and worst offending files, save the result as a tiered migration checklist, and let Cursor or Devin take the next bounded task whenever capacity opens.
Source episode · 27:55
How I AI: Zach Davis's 3 Workflows for Enterprise Engineering with AIwith Claire Vo
Zach pipes a frontend test run to a log, asks Claude to quantify roughly 1,200 noisy lines, stores the ranked cleanup tasks under agents/migrations, and assigns the next tier to Cursor.
Before you start
What you need
- A repeatable test command
- A captured test log
- Claude Code or another analysis model
- Cursor, Devin, or another coding agent
What you’ll make
A tiered Markdown migration checklist with quantified warning categories, affected files, and small tasks agents can complete independently.
Tools used
- CursorVisitCursor
AI-first code editor
- ClaudeVisitClaude
Anthropic AI assistant
- Devin
AI software engineer by Cognition Labs
Step by step
The workflow
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
5 steps
Step02
Analyze Logs with AI
Give the log to Claude and ask it to group repeated warnings, quantify each category, identify the worst files, and distinguish legitimate defects from harmless but distracting output.
Example prompt
Analyze this test log as a technical debt inventory. Count repeated warning signatures, group them by root cause, identify the worst offending files, and separate likely correctness problems from cleanup noise. Cite representative log lines. Return a tiered Markdown checklist where each task has scope, affected files, expected line reduction, and a validation command.
Step03
Create a Prioritized Task List
Save the analysis under a migration directory in the repository. Keep the instructions and checklist together so any coding agent can choose one unchecked task and understand the baseline and validation.
Step04
Assign Tasks to AI Agents
Assign one tier or one small task to Cursor or Devin. Require the agent to fix the cause, keep the diff within the named files, run the relevant tests, and report the before and after warning count.
Example prompt
Take the next unchecked task in `agents/migrations/test-noise.md`. Fix the underlying cause without blanket warning suppression. Keep the change within the listed scope, run the documented validation command, update the before and after count, and check off the item only if the tests pass.
Step05
Review and Merge Changes
Review the focused change, compare test counts and warning output with the baseline, and merge it through the normal path. Keep the checklist updated so another agent can take the next item without repeating analysis.
What good looks like
- The baseline records total noise and warning categories.
- Each checklist task names files and a measurable expected reduction.
- One agent can complete a task without taking the entire migration.
- The test log shrinks while tests continue to pass.
Build your next product with ChatPRD
Turn an idea into a PRD, user stories, and a plan.
After the steps
Runbook notes
How to recover when the loop fails and where human judgment helps.
Recover
If it goes sideways
- The log is too large for one model request
- Have the agent process the file in chunks, aggregate counts by warning signature, and cite sample lines and affected files in the final analysis.
- The agent hides warnings instead of fixing causes
- Forbid blanket suppression and require the task to explain the root cause and before or after count.
- A cleanup task touches too many files
- Split it by warning type or small file group and run the test command after each branch.
- Noise drops because tests stopped exercising behavior
- Compare test counts and failures with the baseline, and inspect any deleted assertion, mock, or setup code.
Keep building
Related workflows
advancedPersonalClaude
Create a Meta-Workflow to Continuously Improve Your AI Assistant's Performance
Design a weekly meta-workflow that analyzes your interactions with your AI assistant to make it smarter. This system learns your writing style from edits, suggests new automations for repetitive tasks, and helps you filter valuable AI techniques from hype.
Open
intermediateProductClaudeGranola
Build a Self-Improving AI Morning Brief to Capture Action Items and Learn Company Jargon
Create a daily AI workflow that summarizes yesterday's meetings, identifies urgent tasks, and proactively learns internal company terms it doesn't understand. This keeps your AI assistant's knowledge current and highly relevant to your work.
Open
intermediateProductClaudeNotion+1
Automate Your Weekly Planning with an AI-Powered PM Assistant
Build a recurring AI task that automatically gathers updates from your calendar, Slack, and notes to suggest weekly priorities and prepare you for upcoming meetings. Start every Monday with a clear, comprehensive plan without the manual effort.
Open
Start shipping
better products.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.
Free to start
No credit card
SOC 2 certified
Enterprise ready