Set up Claude Code with project rules and logging

A developer paired with Claude Code on a Rust app, using a rules file and precise logs to guide planning, coding, testing, and debugging.

Done with Claude Code· Intermediate· Oct 2026

What they did
They spent roughly seven months building ThinkTerm, a Rust terminal multiplexer built on WezTerm, with Claude Code acting as a partner on the work. The agent took on planning, writing code, running tests, and checking benchmarks, while the developer reviewed what it produced. They kept the important constraints in AGENTS.md or Claude.md, and when they needed interface work without a framework, they had Claude export window frames as PNGs because it could not directly inspect the display.
How it went
The finished project is ThinkTerm, an open-source app for macOS, Linux, and Windows that the developer uses to organize agent sessions across machines. The biggest weakness they mention is debugging: Claude kept choosing the wrong explanation for one bug until they changed the program to record actual numeric dimensions. That extra logging solved the problem after five incorrect guesses.
Tips
Put the project rules in a file the agent will read before it starts, and require it to report real numbers rather than inferred values. If a task could send, delete, submit, or otherwise commit something irreversible, make the agent wait for your approval first.

Try it yourself with Claude Code

Use [Claude Code] to help with [feature or bug] in [repo path]. First, read the project rules from [AGENTS.md or Claude.md], then outline a plan, make the changes, run the relevant tests or benchmarks, and inspect the output for problems. If you need to send, delete, submit, or change anything irreversible, pause and ask me to confirm before doing it. When you debug, add numbered log statements that print exact measurements and other concrete values, not guesses, so you can verify behavior from the logs. When everything is working, give me a short summary of what changed and anything I should watch out for.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on reddit.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.