Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Reduce max papers per expert from 6 to 5 (3-5 optimal to avoid context contamination); add Anti-Fabrication Rule (leave blank if source not found, never fabricate citations); add No Source = No Notes policy
What This Skill Does
Orchestrates multiple AI agents to simulate an academic reading group across 1-50 papers. Agents read papers in parallel, a junior researcher poses cross-paper questions, experts respond and debate, then a final agent synthesizes everything into a cited discussion document. All claims are traceable to source passages.
Rather than summarizing papers individually, it surfaces cross-paper tensions and open questions that a single-pass read would miss.
When to use it
- Reviewing a batch of papers before writing a literature review
- Preparing for a thesis committee meeting by stress-testing interpretations
- Identifying contradictions across papers from different research groups
- Running a lab journal club without scheduling conflicts
- Synthesizing a research area you're entering for the first time
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Analyze these 8 PDFs on transformer attention mechanisms through the lens of computational efficiency
- 1Validate inputs: research question, paper list, output directory
- 2Assign 2 experts, 4 papers each; spawn parallel Phase 1 reader agents
- 3Junior researcher agent reads all notes and posts challenging questions to each expert
- 4Spawn parallel Phase 3 expert response agents; each expert replies to junior's questions and engages with the other expert's perspective
- 5Synthesis agent reads all outputs and produces Integrated_Discussion_Summary.md organized by theme with full citations
Integrated_Discussion_Summary.md covering points of consensus, disagreement, open questions, and traceable citations in (AuthorYear, §Section) format