Batch refactoring million-line codebases with Claude

After tightening compiler settings in a mono-repo with 50 projects and over a million lines of code, leverage Claude combined with custom skills or scripts to batch-fix compilation errors. Successfully handled over 20K lines of code modifications and completed the .dotnet 10 LTS upgrade.

agentusecases.ai

Claude

Added 2026-11-03 · originally published Mon, 02 Nov 2026 09:00:00 GMT

Steps

1Enable stricter compiler settings in the codebase to trigger compilation errors.
2Categorize refactoring tasks by logic, performance, and appearance; keep business logic changes in separate sessions and run batch operations only for performance and appearance changes.
3Have the AI compile the project and group errors by type, prioritizing the most common error types.
4Write reusable scripts or skills to safely fix entire classes of errors, recompile after applying modifications to confirm the error count drops and tests pass.
5Repeat the process for other error types until the build is clean or a specified number of rounds is reached, and present a diff summary before committing.

Copy-ready prompt

In the codebase at [path], I want to enable the stricter compiler setting [flag or warning level]. Build the project, group the resulting errors by type, and pick the most common type first. Write a reusable script or skill that fixes that whole class of error safely, apply it, and rebuild to confirm the error count dropped and tests still pass. Repeat for the next error type, and stop when the build is clean or you've done [number] rounds. Show me the diff summary before you commit anything.

View original post

#Code Refactoring#Compiler#Software Development#Automation

Run parallel Claude sessions to build and test apps

Developers maintain about 10 codebase copies with independent Claude sessions, combining strict planning specs and adb automation to drive phones for UI testing and log debugging to rapidly advance app development. Once configured, Claude reportedly executes tasks at very high speeds.

Share logs across multi-AI agents using Markdown files and hooks

To prevent different AI agents from forgetting each other's actions, the author built a collaboration framework using a Markdown file, two hooks, and a configuration file. Over 7 days, the AI agents wrote 1,556 lines to the shared log without human intervention.

Automating two years of overdue taxes with Claude

The user built an automated accounting system around Claude to clear two years of transaction records across three banks, matching receipts and invoices while generating reports in the format required by the Dutch tax authorities. This approach cleared a two-year backlog in a single afternoon and compiled a list of edge cases for the accountant.

Found this useful? Get daily updates on Telegram or X.

Join on Telegram Follow @CaiXiang23766

Found this useful? Buy me a coffee ☕

nebu.life is free to use. If it helped you, a small tip keeps it going. Payments go straight to the author's personal account; this site never handles any funds.

Summarized by AI from the original source - refer to the original post for details. Copyright belongs to the original author.

Steps

Related use cases