What problem does it solve? Running a full AI research project — from literature survey through experiments to a publishable paper — requires constant human supervision, context tracking, and coordination across many tools. This Skill lets an AI agent manage that entire lifecycle autonomously, maintaining structured state and continuously making research progress without waiting for human input. ## Core Features & Use Cases - Two-Loop Research Engine: An inner loop runs rapid, measurable experiment iterations while an outer loop periodically synthesizes results, identifies patterns, and steers research direction (deepen, broaden, pivot, or conclude). - Structured Research Workspace: Initializes and maintains research-state.yaml, research-log.md, findings.md, literature notes, and per-hypothesis experiment directories so context survives across sessions. - Domain Skill Routing: Maps research activities (training, evaluation, interpretability, inference, MLOps) to the appropriate domain-specific skills in the library for execution. - Continuous Operation: Sets up a 20-minute wall-clock loop via Claude Code /loop or OpenClaw cron so research never stalls, plus HTML/PDF progress presentations for human review. - Use Case: Give the agent a question like "do sleep-like consolidation phases accelerate grokking in small transformers" — it surveys the literature, forms hypotheses, runs experiments, synthesizes findings, and ultimately drafts a paper via the ml-paper-writing skill. ## Quick Start Start an autonomous research project on my question about efficient transformer architectures, set up the workspace and continuity loop, and keep me updated with progress reports.
