Using AI in the instro lib to add support for new instruments

The instro repository includes reusable skills and subagents for both Claude (instro/.claude/) and Codex (instro/.codex/). These packaged capabilities enable your agents to add support for new instruments and run validation tests if you have the physical device connected to your computer.

In the video below, I show how I used the add-instrument-driver and validate-driver-hardware skills with Claude Code to add support for the Keysight 34461A Digital Multimeter in the instro library. With the DMM connected to my computer over USB and a few simple prompts in the terminal, Claude was able to add a new InstroDMM driver for the Keysight 34461A, run through a series of software and hardware tests, and open a PR in the github repo - all in under 25 minutes.

What used to take me several days to write Python code to command a VISA/SCPI benchtop instrument and perform validation tests now takes me less than a half an hour with the power of AI and the structure of the instro library.

Nice workflow. One thing I’d want in the validation skill is a captured trace of the exact SCPI commands and responses, so reviewers can diff behavior across firmware versions. Hardware-in-the-loop catches timing and connection issues, but a golden trace makes protocol regressions visible even when the instrument isn’t attached.