web-design-guidelines
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external source code during the review process, which creates a potential surface for instructions embedded within that data (such as in code comments) to influence the agent's behavior.
- Ingestion points: The skill identifies and reads project UI source files, components, and styling rules as specified in the workflow of SKILL.md.
- Boundary markers: There are no explicit boundary markers or instructions to the agent to treat content from reviewed files as untrusted data or to disregard instructions found within comments or metadata.
- Capability inventory: The skill utilizes file system tools (Read, Grep), can be configured for a "fix-loop" involving file modifications, and has authorization to use subagents via the
spawn_agentcommand. - Sanitization: No sanitization, escaping, or validation of the untrusted code content is performed before the agent processes it for review.
Audit Metadata