What is ClawHub? The OpenClaw skills registry, and how to read its security audits

ClawHub is the public registry for OpenClaw skills and plugins, the place the openclaw skills search and install commands resolve to. It is run by the OpenClaw project, lives at clawhub.ai, and it tracks versions, changelogs, downloads, stars and a security scan for every listing. The registry answers two questions: what can my agent learn to do, and should I trust this bundle of files to teach it.

What it hosts

• Skills: versioned text bundles built around a SKILL.md file, with supporting files, templates and scripts.
• Code plugins: packaged OpenClaw extensions with compatibility metadata, installed with an explicit registry source.
• Bundle plugins: packaged plugin bundles for OpenClaw distribution.

Publishing is open. Anyone can push a version with the separate clawhub CLI, and each version is an immutable record with its files, source attribution and scan status. Installing goes through the native openclaw commands, which record where a skill came from so later updates stay on the registry, and plugin installs check the advertised gateway compatibility before anything runs.

Reading a security audit

Every listing shows an audit status and a risk level, and they answer different questions. The status says what to do: Pass, Review, Warn, Malicious, Pending or Error. The risk level says how much power the release has if used as intended: Low, Medium or High. A publishing skill can be Review with Medium risk and still be legitimate; it means purpose-aligned, with meaningful account authority.

Audits are strong safety signals, but they are not a guarantee that a release is risk-free.

• Check the status, the risk level and the listed findings, each of which says what it means and what to do.
• Read the required credentials, permissions and environment variables before you grant them.
• Weigh owner, source, version, changelog, downloads and stars as trust signals, not as proofs.

Why an open registry needs a second reading

Open publishing is the point and the problem. ClawHub applies upload gates, automated checks, user reports and moderator action, and held or blocked content disappears from search. That still leaves a large middle: thousands of skills whose scan has not finished, or whose Pass only means nothing above low risk was visible. One of the most installed skills on the registry is a security checklist for vetting other skills, which tells you what its users think of the middle.

How Diali reads it for you

Diali’s skills catalog mirrors the registry for both engines and runs its own review on each skill: a verdict, a plain-language summary of what the files actually do, and the date it was checked, shown before you install. The skill runs inside the assistant’s sandbox, so one that misbehaves reaches nothing else. The Diali skills catalog is the catalog, and every skill page carries the review.

In short

• ClawHub is the registry; openclaw installs from it and clawhub publishes to it.
• Status says what to do; risk level says how much power is at stake.
• A Pass is reassuring and not a guarantee; read the findings for anything that touches accounts, files or money.

For a starting list that has been through both readings, see the best OpenClaw skills to install first.