A sufficiently decentralized social network. farcaster.xyz

we now have miniapps.farcaster.xyz/llms-full.txt and `pnpm create @farcaster/mini-app`

what's the best way to integrate these two? when we setup a starter repo should we have a claude.md? what about other agents? what should go in that setup file?

Building @warpcast and @farcaster / dad / like to read, cook, ski, code

love!! what i do is make a docs/frame_integration.md (or mdc i’m never sure) file and then in the CLAUDE.md file say “look at docs/ for step by step instructions”

not fully following, what's in docs/frame_integration.md vs what's in claude.md?

the full docs you have for frame would go in docs/ — claude.md id just the default file claude code looks at so i have “look at the docs instead” in it + some other stuff like “never do anything i don’t tell you to do” and etc — you COULD put the entire thing in claude.md i guess but people might want to own that themselves

it really depends on the use case i think so i'd let them do it themselves (+ not everyone is going to be using Claude Code) -- most of my projects are just stuff like "NEVER USE A - IN ONCHAIN" and "DO NOT SAY IN THE README ANYTHING IS SIMPLE NONE OF THIS SHIT IS"

last question do you have opinionated take on what meta instructions are good for claude.md? or is this important for developers to arrive at themselves

by default it's not gonna look for it unless you tell it to, or at least that's been my experience -- tbh i don't use claude.md i just start every chat with "Look at docs/.. and do ..."

got it, so we short put the entire llms-full.txt file in docs

can claude not find it by at the live url