artifact-publish
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides explicit warnings against publishing sensitive content, credentials, or secrets without user confirmation. It emphasizes that while Artifacts are default-private, they are hosted on an external service (claude.ai) and should be treated with the same caution as any other outward-facing action.
- [SAFE]: The technical instructions correctly identify that the platform enforces a strict Content Security Policy (CSP) that blocks all external requests (CDNs, fonts, images, scripts, fetch/XHR). It mandates inlining all CSS and JS and embedding assets as data URIs, which prevents common cross-site scripting (XSS) and data exfiltration vectors.
- [SAFE]: The tool configuration (allowed-tools) is restricted to necessary operations (Read, Write, Edit, Artifact, WebFetch) and does not request unnecessary permissions like shell access or direct network requests outside of the Artifact infrastructure.
- [SAFE]: The skill implements conflict resolution (409 handling) and encourages stable favicons and titles to prevent user confusion during redeploys.
Audit Metadata