SciSkills 🧪🤖
Scientific Embodied Skill Library for AI Agents
Building blocks for
autonomous scientists.
SciSkills provides a curated library of atomic and composite laboratory skills, fully interoperable via the Scientific Context Protocol (SCP). Designed for PLAD agents, LabVLA, and embodied AI systems.
Unified Description
Every skill is defined with four core components:
- 📋 Preconditions – required instrument states, calibration, environmental conditions
- ⚙️ Operation Parameters – volume, temperature, speed, duration, tolerances
- 🔒 Safety Constraints – pressure limits, chemical compatibility, emergency stop conditions
- 📊 Expected Outputs – measured data, state changes, success criteria
SCP Integration
Scientific Context Protocol standardizes how skills are exposed, discovered, and orchestrated. Agents can query available skills, validate preconditions, and execute them across heterogeneous lab infrastructure.
{
"skill_id": "pipette.liquid_transfer",
"preconditions": {"tip_attached": true, "volume_range": [1,1000]},
"parameters": {"volume_ul": 250, "source": "A1", "target": "B3"},
"safety": {"max_pressure": 0.2, "chemical_group": "acid"},
"output": {"actual_volume": "float", "error_code": "int"} }
Liquid Transfer (Pipetting)
Transfer precise volume between vessels with tip ejection and air displacement control.
🎥 Skill demo (3D animation / video preview) — reserved
Interactive visualization coming soonHeating & Stirring
Control hotplate/stirrer with ramp rate, target temp, and stirring speed under inert atmosphere.
🎥 3D animation / sensor overlay — reserved
Microscopy Imaging
Automated focus, stage movement, multi-channel acquisition with image quality assessment.
🔬 live demo / simulation — reserved
More than 150+ atomic skills and 40+ composite workflows available. Full skill browser with filters, versioning, and agent API coming in Q1 2027.
Agent-ready skill integration
SciSkills is natively compatible with PLAD agents, LabVLA, and any SCP-enabled robotic system. Skills can be dynamically queried, composed into experimental protocols, and executed with full safety traceability.