Files
UniversalTrain-ai/AGENTS.md
T
chenlin c32df70326
Checks / Layout, translation and Colab checks (push) Has been cancelled
Checks / Ruff (push) Has been cancelled
Checks / Framework tests (native backend, CPU) (push) Has been cancelled
Checks / Portable Rust controller (push) Has been cancelled
release: v0.1 generic robot training and application platform
2026-11-05 11:05:01 +08:00

2.5 KiB

Repository instructions

Read CLAUDE.md for working rules and Workflow to choose the robot intake route. The repository contains onboarding, native control, simulation training, and packaging. Use the local sources and the asset's supplied documentation to establish its model and control assumptions.

Use lightweight models for bounded reading and documentation, standard models for routine changes, and advanced models for architecture or difficult debugging. Delegate independent tasks when appropriate and preserve validation.

Work Reference
Commands and configuration Usage
Environment setup or repair Agent setup
Model units, frames, collision, and calibration Hardware
Controller and input bindings Controls
Application build or policy deployment Deploy guide and bundle format
Repository conventions and checks Contributing

Skills

Read an applicable SKILL.md in full before starting the work.

Skill Use for
setup-env Establish or repair the Python and controller build environment.
onboard-asset Inspect assets, normalize geometry, choose an intake tier, and verify physics.
tier1-adapt Adapt an existing policy from declarative metadata and check numerical parity.
tier2-control Configure and verify deterministic robot controls.
tier3-train Calibrate and train a robot using a locked locomotion template.
new-task Create the data-only robot specification and verify discovery.
train Run or resume training and report measured replay results.
controls Bind keys and pad controls using the shared vocabulary.
deploy Export and package policies for the supported host.
bundle-manual Translate the built application's operator manual.

Every new skill belongs in this table. The template directories carry their own lock instructions. Onboarding writes one literal robot_spec.py per robot and keeps the calibrated asset, controller, reports, and controls outside the locked templates.