This is the top-level orchestrator for building an entire research paper from scratch. It walks you through all 11 phases (literature search through final PDF compilation), manages dependencies between steps, and saves checkpoints so you can resume if something breaks. The assembly checker script scans your paper directory to figure out what's done and what's missing. It's borrowed heavily from AI-Scientist's sequential pipeline but adds AgentLaboratory-style state propagation between phases. The main value is enforcing the dependency graph so you don't accidentally write results sections before running experiments, and making sure every number in your paper traces back to actual logs. Honestly most useful if you're doing this kind of paper generation repeatedly and want guardrails.
npx skills add https://github.com/lingzhi227/agent-research-skills --skill paper-assembly