Today I launched and pushed the first version of [Better CLI](https://github.com/yogin16/better-cli).

It’s a simple skill that helps build command-line tools that are easier for people to use and easier for AI agents to run.

I made it for a very practical reason: a lot of CLI tools work, but still feel confusing. Errors are vague, help text is thin, and automation breaks on small surprises.

Make CLIs such that it helps AI Agent navigate, instead of fighting with it. Inspiration came, when I was writing CLI for [lore](https://github.com/lorelang/lore), and couldn't find any skill for CLI writing that I was looking for.

So this project is my attempt to keep things straightforward: - clearer commands - clearer errors - output that is easier to reuse - fewer “guess what I meant” moments

To test this skill, I ran it on recently launched Google's [gws cli](https://github.com/googleworkspace/cli):

```bash npx skills add yogin16/better-cli ```

Then on Claude just one prompt: ```bash use better-cli skill and improve the cli as needed with best practices, no regression, ultrathink. ```

The gws cli is already nicely written for AI Agents, still it came up with [these changes](https://github.com/googleworkspace/cli/pull/297/changes) to improve error messages so that AI Agents using the cli knows what do with it.

If you build CLIs (or suffer through them every day), I’d love your feedback. CLI is what the new software is becoming, at least for now. (API -> MCP -> CLIs) Hope this helps you better the CLIs for your software for agents.