Playlist "openSUSE Conference 2026"
amake: a task runner for your AI coding tools
If you've been using AI coding assistants from the terminal (things like Claude Code, Aider, or GitHub Copilot) you've probably noticed that you end up typing the same prompts over and over. "Create a PR for this branch." "Review the diff." "Refactor this module." And when you want to chain them together, you're copy-pasting output between invocations and hoping you don't mess up the order.
I got tired of that, so I built amake.
The idea is simple: you write an Amakefile (yes, like a Makefile, but for AI prompts) where you define tasks. Each task has a prompt, a tool to run it with, and optionally some dependencies. Then amake run review figures out the dependency graph, runs things in the right order, and pipes captured output from one task into the next. It's make for the age of AI-assisted development, same old idea of declarative task dependencies, just pointed at a different kind of tool.
What makes it interesting in practice is that it doesn't care which AI tool you use. Claude Code, Aider, Copilot — they all have different CLIs with different flags and different ways of passing prompts. amake has an adapter layer that handles the translation, so you write the prompt once and pick the tool. You can even mix tools in the same pipeline: have one tool create a PR, then have another tool review it. Or use a completely unknown binary: amake will just shell out to it and pass your prompt as an argument.
Some features:
- template system for variable interpolation ({{vars.focus}}, {{env.HOME}}, {{tasks.describe.stdout}});
- an --edit-var flag that drops you into your $EDITOR when you need to write something longer than a one-liner;
- a dry-run mode so you can verify your dependencies and resolved prompts before anything actually runs.
In this talk I'll walk through the motivation and show some real workflows... and of course it's available in openSUSE!
Licensed to the public under https://creativecommons.org/licenses/by-sa/4.0/