agent-cli-first

A CLI-first development workflow treats the command line interface as the primary contract, ensuring features are built as commands with machine-readable output and semantic exit codes before adding UI or API layers.

To build a CLI that agents can call reliably, you define a clear CLI specification using `--help`, provide machine-readable output with `--json`, and implement semantic exit codes for deterministic behavior.

You make command line tools testable for CI/CD pipelines by exposing core logic as commands first, enabling testable execution, and integrating debugging and profiling paths directly into the development loop.

Use a CLI-first approach for a new microservice when you need core logic to be immediately testable and debuggable by other agents or CI/CD pipelines before building external UI or API adapters.

The best way to structure CLI output for automated workflows is to enforce machine-readable output formats like `--json` alongside semantic exit codes, ensuring deterministic and agent-ready behavior.

Refactoring to a CLI-first contract focuses on exposing core logic via commands first, ensuring features are testable and agent-ready, which can be done before refactoring or building external API adapters.