Security audit
Security checks for vulnerabilities and agentic risk
This documentation-only skill coherently explains Docker sandbox usage, with the main caution that mounted project files and saved sandbox state can still be changed by commands the user runs.
Before using this skill, run untrusted code only against a disposable or backed-up workspace, review network allowlists carefully, and treat reset/remove commands as destructive to sandbox state. The skill itself does not install or run code automatically.
The skill emphasizes VM isolation, but the workspace is mounted from the host via virtiofs, meaning commands in the sandbox can modify or delete host project files. Without a clear warning, users may incorrectly assume the environment is fully isolated and run untrusted code against sensitive host-mounted data.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
Node.js fetch doesn't use the proxy. Use the proxy-fix.js require hook above, or use curl instead:
docker sandbox exec my-sandbox curl -sL https://api.example.com/data
The documented docker sandbox reset command is destructive and clears all sandbox state, but the troubleshooting entry does not explicitly warn about data loss before suggesting it as a fix. Users may run it reflexively during troubleshooting and unintentionally destroy sandboxes, snapshots, or work in progress.
No suspicious patterns detected.