# Nooa

> Build, modify, debug, or optimize agents using NVIDIA-labs OO Agents (NOOA). Use for nooa.Agent subclasses, ellipsis generation methods, CodeAct or Predict strategies, ShellTools, Skill/TextSkill, context and persistence, MCP, tracing, middleware, channels, or trace analysis.

- Skill: `nvidia-nemo/nooa` (Agent Skill)
- Install (CLI): `npx skillmds@latest add nvidia-nemo/nooa`
- Raw SKILL.md: https://api.skillmd.com/api/skills/nvidia-nemo/nooa/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: NVIDIA NeMo (https://skillmd.com/u/nvidia-nemo)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/nvidia-nemo/nooa

---

<!-- SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. -->
<!-- SPDX-License-Identifier: Apache-2.0 -->

# NVIDIA-labs OO Agents (NOOA)

Use the `nooa` package and namespace. It comes from PyPI, and the plugin's own
`pyproject.toml` declares the version floor.

For detailed framework guidance, consult the skills at the upstream tag that
matches the floor:
https://github.com/NVIDIA-NeMo/labs-OO-Agents/tree/v0.0.9/skills

Before changing framework behavior, inspect the installed implementation or the
matching upstream skill instead of guessing. In Experimentalist:

- Call `super().__init__()` before attaching tools or skills.
- Use `GuardedShellTools` where workspace access must remain restricted.
- Hide internal skill registries with `spec(self, "skills", hidden=True)`.
- Keep the explicit `orjson==3.11.9` workaround required by nooa's LiteLLM
  tool-call path.

## Verification

Use the repository root `.venv` and the README's standard `uv` workflow. Run:

```bash
uv run --frozen pytest -q <focused migration tests>
uv run --frozen ruff check .
uv run --frozen ruff format --check .
uv run --frozen pyright
```

Then run the changed agent end to end with its configured model.