open-design-publish
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes and publishes user-provided design artifacts (HTML, CSS, JS, and project files) to external hosting providers. This creates an attack surface where malicious instructions or code embedded within a design artifact could be executed or processed by the agent or the end-user's browser upon deployment.
- Ingestion points: The skill reads design files from the
.open-design/directory and other artifact-related files. - Boundary markers: The skill does not implement explicit delimiters for the design content itself but mandates a human-in-the-loop confirmation step ('stop for the user's explicit yes') before deployment.
- Capability inventory: The skill utilizes tools for file system access (
get_open_design_artifact), file writing (export_open_design_artifact), and network deployment (publish_open_design_artifact). - Sanitization: The skill relies on the underlying tools and platform safety filters; no explicit sanitization of design source code is performed within the instructions.
Audit Metadata