OSM Data Prep — Labelled Recordings → Role Files

The OSM agent learns operational states (drilling / tripping, fill / wash / spin / drain, …) from labelled recordings and classifies every window of new ones. Before anything reaches the platform, the recordings have to become role files: a small, strict layout the Optimizations and Evals APIs accept. Most ways to get it wrong don't error: an optimization with too many training files never starts, one scored window across a time gap fails the whole trial, and a state missing from the test set scores 0. This skill builds the layout and checks every rule before you upload.

recordings (any rate, gaps, jitter)          role files (the contract)
  + a per-row state label        prepare.py   ─┐
  + recordings.csv (index)   →   check_states │→  build_roles.py  →  preflight_roles.py  →  OSM agent skill, Path 2
                                 split_roles  ─┘