vercel-react-native-skills
Fail
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits metadata poisoning by impersonating a trusted organization. The skill title ('vercel-react-native-skills') and the author field in
SKILL.md('vercel') are intentionally deceptive, as the actual developer identified in the author context is 'the-utopia-studio'. This misrepresentation is a tactic to gain unearned trust from users and automated systems regarding the provenance and safety of the skill. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection attacks because it lacks boundary definitions for the external data it processes.
- Ingestion points: The agent ingests third-party React Native source code, project configurations, and file contents as part of its primary functions (refactoring and optimization).
- Boundary markers: The instructions contain no explicit delimiters or specific directives to the agent to disregard instructions potentially embedded within code comments, string literals, or README files of the projects being analyzed.
- Capability inventory: The agent context typically includes capabilities for file system modification, command-line execution, and potentially tool-based network access, which could be abused if the agent interprets data as instructions.
- Sanitization: The skill does not define any sanitization, validation, or escaping logic for the external code content it evaluates.
Recommendations
- AI detected serious security threats
Audit Metadata