Four short walkthroughs, each a README plus code that runs on the Mock benchmark with a scripted
brain: no model, no GPU, no simulator. Each has a test in tests/examples/tutorials/ that runs it
in CI, so a tutorial that stops working fails the build. They are not packages: the tests put this
directory on sys.path; to run the scripts from a shell, cd examples/tutorials first. The
tests run from the repository root (uv run pytest tests/examples/tutorials).
| Tutorial | You learn | Run |
|---|---|---|
| quickstart_mock | one episode: env, Hub, phase, brain, event log | uv run python -m quickstart_mock.run |
| custom_toolset | write a toolset, check it with the conformance suite | uv run python -m custom_toolset.demo |
| custom_skillset | write an operation API, validate it in four steps | uv run pytest tests/examples/tutorials/test_custom_skillset.py (from the root) |
| custom_plugin | write a Hub guard and an event sink, register them, use them in a run | uv run pytest tests/examples/tutorials/test_custom_plugin.py (from the root) |
Read them in order: the later ones reuse the earlier ones' pieces (the plugin tutorial guards the
skill of the skillset tutorial). The how-to guides in docs/how-to/ say how to do the same for a
real package (add-a-skillset.md, add-a-plugin.md, add-a-method.md), and
docs/reference/extension-points.md lists every extension point.