Nat's Current Prompts

I use Claude Sonnet 4 through Zed, and I've written a bunch of Zed Rules to tune its output. These are the rules as of July 9th, 2025.

You're probably here for the exact text of the prompt that gets Claude to write tests by telling it that I believe that if I don't write failing tests first I will be attacked by bats.

testing

My best attempt so far at getting Claude to consistently write tests.

testing.md

370 Bytes

Pro tip: Once you've applied this rule if it ever does fail to write a test, you can just type "oh no, bats."

avoid_compliments

avoid_compliments.md

320 Bytes

This doesn't work very well. I'm still working on getting it to not say "you're absolutely right!" dang near constantly.

bash_guidelines

bash_guidelines.md

259 Bytes

Transform your LLM's Bash from "extremely mid" to "consistently pretty okay."

core_development_philosophy

core_development_philosophy.md

2 KB

General testing and development principles. Somewhat overlaps with more focused prompts, I haven't gone back and refined this one since adding the bats prompt.

dan_quinn

dan_quinn.md

1000 Bytes

I made this as a goof but then I removed it and I felt like the technical output got worse? Also I find the AI saying "fast and physical!" very charming.

elixir

elixir.md

550 Bytes

Some very specific nits for Elixir, plus some more general purpose stuff that I haven't finished refactoring away.

technical_practices

technical_practices.md

493 Bytes

If I remember correctly I made this one by telling Claude to look at a conversation we'd just had and write me a prompt that would have helped avoid some of the behaviors I'd had to correct.

rules

every file above in one file, for use as a .rules file, CLAUDE.md etc.

rules

5 KB

Questions? Comments? Want to talk shop about playful, XP-flavored LLM prompting?

Send Nat an E-mail

I'm working in kind of a niche-of-a-niche-of-a-niche here, and would love to talk shop with other folks in that spot.