This does exactly what it says: reviews your git diff against main and strips out the telltale signs of AI-generated code. It targets the stuff that makes reviewers groan: overly defensive try/catch blocks in already-validated codepaths, comments explaining obvious things, type safety escape hatches like casting to any, and style inconsistencies. Run it after Claude writes code but before you commit. The premise is straightforward: AI assistants are verbose and defensive by nature, and this attempts to normalize the output to match what a human would actually write. You get a brief summary of changes at the end, no detailed explanations.
npx skills add https://github.com/brianlovin/agent-config --skill deslop