This takes your bloated CLAUDE.md and rewrites it using conditional `<important if>` blocks to fight instruction dilution. The problem: Claude Code tells the model that context "may or may not be relevant," so it ignores chunks of your instructions. The fix: wrap domain-specific rules in XML conditions that mirror Claude's own system prompt structure, keeping foundational context bare at the top. It strips out linter territory, deletes stale code snippets, and splits broad rule lists into targeted blocks. You end up with a leaner file where every instruction has an explicit relevance trigger. Worth running if your CLAUDE.md is over 200 lines or you notice Claude ignoring half your guidelines.
npx skills add https://github.com/dexhorthy/slopfiles --skill improve-claude-md