Skip to content

How RunWhen Verifies Skills

Every Skill in the RunWhen Skills Registry is verified by RunWhen engineers before it is published, no matter who wrote it.

Skills (Skill Templates, also called CodeBundles) are the automation that RunWhen AI Assistants run in your environment. Each Skill carries one or more Tools (Tasks): Monitors that run on a schedule and Runbooks that run on demand. Skills are the part of RunWhen that touches your infrastructure, so the registry is curated. Nothing is published to it without passing RunWhen’s verification.

Authorship and verification are separate

Skills are contributed by RunWhen engineers, by partners, and by members of the RunWhen author community. The registry credits whoever wrote each Skill. Verification is RunWhen’s job in every case.

Each Skill Template page in the registry shows both:

Registry fieldWhat it tells you
AuthorWho wrote the Skill: a RunWhen engineer, a partner, a community author, or RunWhen’s AI authoring agent.
SupportRunWhen Supported, for every Skill in the registry.
Security & AccessThe Skill’s access level (for example, read-only), the cloud permissions it needs, and the kinds of data it collects.

A partner or community author is not a lower tier. Every Skill goes through the same process before it is published.

What verification covers

  1. Human review. A RunWhen engineer reviews the Skill’s code, including what it can access and what it can change.
  2. Testing against test applications. The Skill runs in RunWhen test environments against a range of test applications to confirm it produces correct, structured results.
  3. Failure-condition testing. Where applicable, RunWhen engineers use chaos engineering practices to deliberately introduce the failures a Skill is meant to detect, then confirm the Skill detects and reports them. For example, a Skill that diagnoses crash-looping containers is run against a workload made to crash-loop.
  4. Publication. Only a Skill that passes these steps is published to the registry.

Additional review for Skills that make changes

Most Skills are read-only: they gather information and never change your systems. Skills with Tools that make changes, such as restarting a workload or scaling a resource, receive additional layers of human review before they are published.

Inside your workspace, which AI Assistants can run those Tools is still up to you. See AI Assistants.

When Skills are re-verified

A Skill is re-verified when its code has a major version change.

What verification does not cover

  • Private CodeCollections. Skills in custom CodeCollections that you create and register in your own runner configuration are not verified by RunWhen. Your team is responsible for reviewing and testing them.
  • Your specific environment. Verification confirms a Skill works as intended in RunWhen’s test environments. Your environment, permissions, and platform versions may differ.

Contractual guarantees

For contributors

If you contribute a Skill, it goes through this process before it appears in the registry. Before you submit, test it yourself using the Testing guide and the Development Checklist. Contributors can earn rewards through the Author Rewards program.