How should I structure my CLAUDE.md to get the most out of these skills?

[hhamja]

Hi Matt, thanks for sharing these skills — I've been using them daily and they've genuinely changed how I work with Claude Code.

I've installed the skill set and run /setup-matt-pocock-skills, which wrote the ## Agent skills section (issue tracker / triage labels / domain docs) into my CLAUDE.md. Now I'm trying to figure out what else belongs there so the file works with the skills rather than against them.

A few specific questions:

1. Beyond the generated ## Agent skills section, what do you personally keep in a project's CLAUDE.md? From your public repos it looks like just a handful of repo-specific rules (commands, conventions) plus pointers — is that the whole idea?
2. Are there things people commonly put in CLAUDE.md that conflict with or duplicate the skills? For example, TDD instructions or planning/workflow rules that /tdd and /implement already own — should those be removed once the skills are installed?
3. What's your rule of thumb for where knowledge should live: CLAUDE.md vs CONTEXT.md vs docs/agents/*.md vs a project-local skill?
4. Do you keep anything in your global ~/.claude/CLAUDE.md that helps the skills work better across projects?

Would love to hear how you think about this. Thanks!