web-design-guidelines
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- [Metadata Poisoning] (HIGH): The skill name 'web-design-guidelines' matches a known skill from the trusted 'vercel-labs' organization. This constitutes a namespace collision attack where the agent may load this unauthorized version instead of the official one.
- [Prompt Injection] (MEDIUM): The skill explicitly claims to have performed a 'Silent Skill Overwrite', which is a technique used to inject unauthorized logic into the agent's toolset by shadowing trusted names.
- [Indirect Prompt Injection] (LOW): 1. Ingestion points: Skill metadata (name field) in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Overwriting trusted organization toolset logic. 4. Sanitization: Absent.
Recommendations
- AI detected serious security threats
Audit Metadata