Runs a full collaboration audit on your Claude setup and tells you what to fix. Exposes collab, security, and health reports that scan your CLAUDE.md or AGENTS.md for leaked secrets, conflicting rules, and stale configurations. Everything stays local unless you generate a Wrapped card and explicitly share it. The collab tool gives you a 0-100 score across seven categories and a persona classification. Security scans for twelve types of secrets including OpenAI, Anthropic, AWS, and GitHub tokens. Health checks for orphaned tasks and overlapping skills. Includes onboard for first-time setup, history for trends without rescanning, and implement_recommendation to apply fixes directly. Reach for this when your agent contract has grown organically and you want an outside view on what's broken or leaking.
claude mcp add --transport stdio bleedmode-dearuser uvx dearuser