It's the AGENTS PR again... @Arthur and @Jonathan Manning are suggesting to remove the treemap from the file and replace it with a link to the template directory in the tools repo. Since the treemap was one of the main features requested in the original discussion, I don't want to remove it without further input. Can we get some more opinions on the issue?

I just think stopping it drifting out of synch will be a nightmare, unless it could be automated

And it also ties the agent to a singular version of tools. It's just not very helpful imo

Makes sense, although IMO we should still provide some key facts (e.g. the concept of local vs. nf-core modules, which is not obvious from a bare treemap)

I think general structure is fine, but I've used Claude a bit with my own nf-core templated pipeline and beyond being told where to find modules/subworkflows given their name it really doesn't seem to need anything like that

Having some automation to generate an AI-readable statement of nf-core standards based on the template somewhere central isn't a bad idea though, just not human-maintained

I'm thinking about removing irrelevant files and reducing the treemap to describe

conf

,

modules

(with

nf-core

and

local

),

subworkflows

,

workflows

,

tests

,

nextflow_schema.json

and

<http://main.nf|main.nf>

. What do you think? Too much, anything commonly relevant missing?

I think that is a good high-level structure to cover

I agree, just list the important files and where they are

I have minified the structure, let me know if it's okay like this