Reading a ClawHub security audit

A security audit on ClawHub exists to answer one question before you install something: what does this release do, and how much authority is it asking for. It is a strong signal and a fast one, and the documentation is refreshingly direct about where the signal stops.

The six statuses

• Pass means no visible issue above low risk was found; Review means read the findings first, because the release may still be perfectly legitimate.
• Warn means a high-impact concern or a warning signal was found and you should use extra caution.
• Malicious means do not install, with no further interpretation required.
• Pending means the audit has not finished, and Error means it could not be completed, neither is a pass, and treating either as one is the most common misreading.

Audits are strong safety signals, but they are not a guarantee that a release is risk-free. Always use judgment before granting sensitive access.

Risk level is blast radius

Status and risk level answer different questions, and conflating them is the second common mistake. Status tells you how to react to what the scanner found. Risk level describes blast radius: how much power the release appears to hold if it turns out to behave badly. A low-risk release with a Review status and a high-risk release with a Pass are very different propositions, and only reading both tells them apart.

What to check before installing

• The overall audit status, the risk level, and any findings that are listed rather than just the headline.
• The credentials, permissions and environment variables the release requires, which is the most concrete statement of what it will be able to reach.
• The trust signals around it: owner, source, version, changelog, downloads and stars. The advice reduces to one line, install only content you understand and trust.

The caveat about a pass matters most for exactly the tools people most want to install: anything that can publish content, edit data, run commands, read files or reach production systems. A scanner reads what the code appears to do; it cannot read what a maintainer will do next. See ClawHub moderation and account standing for what happens when a release is flagged and The ClawHub skill format for where the declared requirements come from.

Getting the report itself

The audit is not only a badge on a page. A publisher can pull the full report archive from the CLI, and a separate download retrieves the stored report for a version that was blocked or hidden, which is precisely when an author most needs to read it. Verification is stricter than the badge suggests: a release verifies as ok only when it has a generated Skill Card, is not blocked as malware by moderation, and carries a clean scan verdict. See The ClawHub CLI for the commands and The ClawHub HTTP endpoints for the endpoint.

On Diali

Diali hosts OpenClaw, and an audited skill still runs with whatever authority your assistant grants it. Each customer runs their own assistant on isolated infrastructure, the runtime configuration is generated from the dashboard and replaced at each release, and state lives on a persistent volume, with daily snapshots and one-click restore available through the Backups add-on (included on Max). See Hosted OpenClaw on Diali for what the hosting includes and Diali pricing for what it costs.

• Pending and Error are not passes.
• Status is how to react; risk level is how much is at stake.
• Read the required credentials before the star count.