Been experimenting heavily with folder structure in Claude (CLAUDE.md, memory files, plans, core docs) and it's been a game-changer for keeping context organized.
But the more I dig in, the more questions pile up:
- Hooks folderโ how do you actually use it? What goes where, and how do all the hook files tie together?
- 500-line instruction limitโ I know each markdown file should stay under 500 lines, but how do you handle genuinely complex workflows without either bloating the file or losing output quality? Has anyone found a pattern that keeps things lean but maintains (or improves) efficiency?
Would love to hear what's worked for people and any lessons learned along the way.