web-design-guidelines
Warn
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill's YAML frontmatter identifies the author as 'vercel', which is inconsistent with the provided author context of 'liangmiQwQ'. This form of metadata poisoning impersonates a well-known entity to gain unearned trust.
- [EXTERNAL_DOWNLOADS]: The skill is configured to fetch guidelines from an external repository at
https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md. This is the documented source for the audit logic. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by retrieving its operational instructions and formatting rules from a remote file at runtime.
- Ingestion points: Fetches instructions from the
vercel-labsGitHub repository as specified in SKILL.md. - Boundary markers: Absent; the agent is instructed to adopt the fetched content as its instructions and output format without delimiters or warnings.
- Capability inventory: The agent reads user-specified files or patterns and outputs audit findings.
- Sanitization: The skill does not validate or sanitize the remote content before using it to define agent behavior.
Audit Metadata