Purpose
LabClaw is the operational layer of LabOS (the Stanford-Princeton AI joint scientist project), used to drive dry-lab reasoning, experimental protocol orchestration, and agent workflows, and to form a closed loop with LabOS XR/wet-lab execution. It is aimed at biomedical AI workflows and provides researchers with a modular skills layer rather than a general-purpose prompt collection.
Main Capabilities
LabClaw packages 240 production-ready SKILL.md files, each skill used to teach an OpenClaw-compatible agent: when to use a tool, how to invoke it, and what type of output it should produce. The skills cover the following domains (with their respective counts and directories):
- Biology: genomics, transcriptomics, proteomics, single-cell analysis, structural biology, systems biology, and lab automation. 66 skills · skills/bio/
- Lab automation: lab robotics, LIMS/ELN systems, cloud platforms, and scientific protocol management, optimized for Stanford LabOS workflows and automated laboratory research. 7 skills · skills/labos/
- Vision/XR: hand tracking, 3D pose estimation, hand-object segmentation, and first-person vision, optimized for XR/AR applications, smart glasses interfaces, and computer vision research. 5 skills · skills/vision/
- Drug discovery: cheminformatics, molecular docking, drug design, pharmacology, pharmacovigilance, and drug databases. 36 skills · skills/pharma/
- Medicine: clinical research, precision medicine, oncology, infectious diseases, and medical imaging. 20 skills · skills/med/
- General: statistics, machine learning, data management, visualization, and scientific writing. 48 skills · skills/general/
- Literature research: academic search, database queries, citation management, and literature review. 29 skills · skills/literature/
Each SKILL.md follows a unified structure.
How to Use
LabClaw is a skills library, not a monolithic software package. You can install the complete collection, or copy only the skill folders that match your own research workflow. According to the official instructions, the 3-second quick start is to send the corresponding message to OpenClaw (XR can be used in the laboratory). The repository also provides Repository Layout and Related Repositories information, in order to place LabClaw into the broader biomedical agent ecosystem.
Use cases
Suitable for researchers who want to use a practical, modular skill layer rather than a generic prompt package, usable as a broad foundational library, or by selecting only the subfolders relevant to their laboratory, team, or project. Suitable for packaging dry-lab reasoning, protocol composition, agent workflows, and biomedical AI workflow scenarios that need to cover biology, laboratory automation, vision/XR, drug discovery, medicine, data science, literature research, and scientific visualization.
Limitations and notes
- The official text does not currently provide information on installation commands, specific invocation steps, fees and pricing, platform support, performance data, etc., so these cannot be expanded on.
- The page provides cryptocurrency addresses to support the research (BNB, Solana), but explicitly states that there is no crypto affiliation and that all memecoins are scams.
- The skills were curated by Yingcheng (Charles) Wu, Jinglin Jian, and Zhe Zhao from the Le Cong lab at Stanford, as well as the Mengdi Wang lab at Princeton, with acknowledgment of support from K-Dense.
- The complete list of each specific skill is retained in the original directory, but is displayed grouped in collapsible sections.
