bbt is the most effective solution for documenting and testing CLI software: describe your scenarios in near-natural language in the format of your Markdown documentation, and then just run this doc.
Version 0.4 focuses on what happens when things go wrong:
- the commands’ standard error is now captured apart from the standard output, and can be checked with
Then the error output is | contains | does not contain ...; - the exact exit code can be checked with
Then the exit code is `1`; - environment variables are now under the scenario’s control
Given the environment variable `LANG` is `C`(or... is not set); - when a check fails, expected and actual outputs are now displayed in a much more comfortable side by side format, sdiff inspired, with a few lines of context.
Special Kudos to @Heziode, who largely contributed to this version’s contents.
For AI coding agents
bbt now ships an agent skill that teaches AI coding agents (Claude Code, Cursor, Copilot, …) to write, convert, run and debug bbt scenarios. Installing it is very simple: see the For AI coding agents section on the bbt site for the instructions.
Then just ask your LLM to make the examples in your README verifiable, or to turn your spec into executable tests, and that’s it!
bbt is available for Linux/Mac/Windows, and has been submitted to the Alire community index.