SciAgent-Skills — Open-Source AI Agent Skill Library for Lab Automation (Python)

General
1

Lab automation scripting with Opentrons, PyLabRobot, or Benchling is powerful — but getting an AI coding agent to generate correct code for these tools is still hit-or-miss. Hallucinated API calls,
wrong parameter names, outdated patterns.

I’ve been building SciAgent-Skills to address this: a collection of structured skill files that AI coding agents (Claude Code, Cursor, etc.) read before generating code. Each skill file contains
runnable examples, key parameters, and troubleshooting guidance — grounded in the actual API docs.

Currently includes 5 lab automation skills:

  • Opentrons Protocol API — OT-2/Flex protocols, module control, local simulation
  • PyLabRobot — hardware-agnostic scripting across Hamilton, Tecan, Opentrons
  • Benchling SDK — registry entities, inventory, ELN entries via Python
  • protocols.io API — programmatic search and retrieval of wet-lab protocols

It’s still early and coverage is thin. I’d love input from people who actually work with these systems — what’s missing, what’s wrong, what would make an AI agent genuinely useful in your automation
workflow.

GitHub: GitHub - jaechang-hits/SciAgent-Skills: Life sciences computational skills for scientific AI agents · GitHub

PRs and issues very welcome.

3 Likes