This orchestrates parallel scientist agents to tackle complex research questions by breaking them into independent stages, running them concurrently, then cross-validating and synthesizing results. You get AUTO mode for fully autonomous execution or checkpointed runs where you approve each phase. The smart routing is the useful bit: it automatically assigns haiku for simple data gathering, sonnet for standard analysis, and opus for deep reasoning across architecture or cross-cutting concerns. Session management persists everything to `.omc/research/` with structured findings and reports. It caps at 20 concurrent agents and includes loop control with promise tags, so you're not babysitting infinite execution. Good for codebase audits, security reviews, or any investigation that splits cleanly into parallel workstreams.
npx skills add https://github.com/yeachan-heo/oh-my-claudecode --skill sciomc