What problem does it solve? Writing Python CLIs with argparse, click, or fire involves boilerplate, manual help text, and fragile argument wiring. This Skill provides conventions and patterns for building CLIs with tyro, where type annotations and docstrings generate the entire interface and --help output automatically. ## Core Features & Use Cases - Type-Driven CLI Generation: Build argument parsers from dataclasses, function signatures, and Literal types with zero argparse boilerplate. - Self-Documenting Help: Module docstrings become program descriptions and field docstrings become argument help, so --help is always complete and current. - Subcommand & Nested Config Patterns: Use SubcommandApp decorators, Union-based dispatch, or nested dataclasses for hierarchical flags like --optimizer.lr. - Machine-Consumable Output: Standardize --json and --plain flags with documented JSON schemas so scripts and LLM agents can consume CLI output reliably. - Use Case: You need a standalone script that processes experiment data. Write a dataclass with typed fields and docstrings, add PEP 723 inline metadata, and run it with uv run script.py --help to get a fully documented CLI. ## Quick Start Ask the AI to create a Python CLI script using tyro with a dataclass for arguments, field docstrings for help text, and PEP 723 inline dependencies so it runs with uv.