Spawns parallel sub-agents to document your codebase as it exists right now, no suggestions attached. You ask a question, it breaks down the research across specialized agents (scout for code exploration, thoughts-locator for historical docs, linear-searcher for tickets), waits for all of them to report back, then synthesizes everything into a markdown research document with full file paths, line numbers, and GitHub permalinks. The whole thing is explicitly built to resist the urge to recommend improvements. It just maps what's there. Useful when you need to understand a complex system quickly or onboard someone without editorializing about what should change.
npx skills add https://github.com/parcadei/continuous-claude-v3 --skill research