web-design-guidelines
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (INFO): The skill fetches guidelines from a remote URL. Evidence:
https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md. This is a trusted source (Vercel organization), which downgrades the network access finding per [TRUST-SCOPE-RULE]. - [REMOTE_CODE_EXECUTION] (LOW): The skill retrieves a remote markdown file (
command.md) containing instructions that the agent is explicitly told to follow. This is a form of remote instruction execution. The risk is mitigated by the trusted status of the repository. - [PROMPT_INJECTION] (MEDIUM): The skill is vulnerable to indirect prompt injection (Category 8) as it processes untrusted user UI files alongside remote instructions.
- Ingestion points: Remote
command.mdguidelines and user-provided local source files. - Boundary markers: Absent; the skill does not use delimiters to separate instructions from data.
- Capability inventory: The skill utilizes
WebFetchfor network requests and can read local files. This combination allows for potential exfiltration if a malicious file manipulates the agent. - Sanitization: None; the content is processed directly by the agent without validation.
Audit Metadata