supervisor-drive

What problem does it solve?

Manually coordinating multiple specialized research agents for academic paper development is time-consuming and error-prone, requiring constant state tracking, task prioritization, and safety checks to avoid unintended changes to research integrity.

Core Features & Use Cases

• Hardcoded bottleneck ranker: Automatically prioritizes the highest-impact research task (drafting unwritten sections, filling missing citations, responding to reviewer feedback, refining figures) based on real-time project state.
• Six mandatory safety gates: Requires explicit user confirmation for high-risk actions (hypothesis changes, new citations, budget overruns, structural rewrites) even in fully autonomous mode, preventing accidental modifications to core research content.
• Three flexible operation modes: Supports interactive mode for step-by-step user approval, auto mode for hands-off autonomous execution, and plan-only mode to preview upcoming actions without making changes.
• Use Case: A researcher finalizing a conference submission can run the skill in auto mode to automatically address all remaining reviewer feedback, fill pending citations, and refine incomplete figures without manually switching between tools or tracking progress.

Quick Start

Invoke the supervisor-drive skill with the /supervisor-drive slash command to start an automated, safety-guarded orchestration loop that coordinates all OMCR research agents to advance your academic paper project.