Add claude.md

[jordanschalm]

Created this using /init focussing on telling claude how to run tests

[Kay-Zee]

does naming it claude.md help direct claude to look here first? if so is the plan to continue to grow this MD file to more than testing?

[vishalchangrani]

does naming it claude.md help direct claude to look here first? if so is the plan to continue to grow this MD file to more than testing?

I have this same question :)

[jordanschalm]

does naming it claude.md help direct claude to look here first?

Yep, CLAUDE.md files are auto-loaded into the context window for Claude Code. So, they are always available, but also always consume context space. The current file is using 2k tokens, which is a lot, so going to trim it down before merging. My main goal here was for all Claude sessions to know how to run individual test cases and test files, which I think can be accomplished with a lot less tokens.

if so is the plan to continue to grow this MD file to more than testing?

Yep, whatever we want to be always loaded into context for the project.

[Gornutz]

does naming it claude.md help direct claude to look here first? if so is the plan to continue to grow this MD file to more than testing?

Yes it helps direct claude directly to look there. If the repo is going to support multiple different agent types (codex / etc) then it's actually better to have it be Agents.md where it's agnostic to the provider. It's also able to provide better context when using skills then having the individual skill available for usage.

Source with benchmarks from Vercel - https://vercel.com/blog/agents-md-outperforms-skills-in-our-agent-evals

[jordanschalm]

Trimmed it to 310 tokens and added an AGENTS.md symlink so other models will pick it up