Skills

Antithesis skills help your AI agents use Antithesis, speeding up the time it takes to set up your system and start finding bugs.

Our skills enable agents to help with every stage of the Antithesis workflow:

• understanding how best to test your software
• setting up your system to run in the Antithesis environment
• writing workloads to exercise your system
• launching test runs
• triaging test results
• debugging

This page introduces the available skills and explains how to install them. For more information, see the antithesis-skills repo.

Requirements

For your agent to use Antithesis, you’ll need:

• Snouty CLI
• Docker or Podman
• npm

These aren’t required if you’re just using antithesis-research, since that skill doesn’t involve Antithesis.

Installing skills

The recommended way to install our skills is to use the npx skills installer:

npx skills add antithesishq/antithesis-skills

Select the following options from the interactive menu:

1. Skills: select the skills you need.
2. Install scope: choose global, not project.
3. Install method: choose symlink.
4. Install find-skills skill: choose No.

After installing, restart any open agent sessions so that they discover the new skills.

To update your skills, run:

npx skills update

Available skills

Teach your agent how to use Antithesis

The antithesis-documentation skill uses snouty docs to read and navigate the docs. It’s used by other skills, or you can use it directly to ask questions about Antithesis.

Understand your system’s properties

The antithesis-research skill analyzes your system and suggests properties your system should have (and that you should test for).

It produces a set of human-readable artifacts which are also used as inputs by the antithesis-setup and antithesis-workload skills.

The antithesis-research skill can be used as a standalone, without Snouty or Antithesis.

Set your system up to test in Antithesis

To set your system up to run in Antithesis, use the antithesis-setup skill if you use Docker Compose, or antithesis-setup-k8s if you use Kubernetes.

They create the config files and instrumentation you need to run your software in Antithesis, then validate it with snouty validate.

These skills expect the output from the antithesis-research skill.

Write a workload

To create a workload for use in Antithesis, use the antithesis-workload skill.

This skill turns properties from your catalog into SDK assertions and test commands, one property at a time, then validates the result with snouty validate.

antithesis-workload expects the output from the antithesis-research skill.

You can also use this skill to refine Antithesis coverage using the output from the antithesis-triage skill.

Launch a test run

When you’re ready to launch your test run, use the antithesis-launch skill.

This skill builds your images, validates your config with snouty validate, and submits the run with snouty launch, stopping if any step fails.

Triage your results

The antithesis-triage skill uses snouty runs to get logs and property failures from a test run.

It investigates property failure examples by fetching logs from around the example moments, then checks system source code to investigate why the failure might have occurred. It then suggests follow-up actions, such as debugging a specific failure or refining your workload.

Debug your software

If analysing the logs isn’t enough to determine what went wrong, use the antithesis-debug skill to dig into a specific failure with Antithesis’ Multiverse Debugger.

The Multiverse Debugger allows your agent (or you) to inspect your container filesystem and runtime state at the moment of failure, run shell commands inside containers, and explore hypothetical execution paths in realtime.