vercel-react-best-practices
Warn
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [METADATA_POISONING]: The skill metadata and documentation (SKILL.md and AGENTS.md) explicitly claim to be authored by 'Vercel Engineering' and 'vercel'. This contradicts the actual author attribution to 'lucasduarteinacio'. This identity spoofing is deceptive and may lead users or AI agents to trust the provided guidance and recommended third-party tools based on a false authority.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to guide AI agents in refactoring and generating code, which creates an attack surface for indirect prompt injection if the agent processes source code containing adversarial instructions.
- Ingestion points: The skill is triggered during tasks involving React/Next.js components, pages, and data fetching (SKILL.md).
- Boundary markers: Absent; there are no instructions for the agent to use delimiters or treat ingested code as untrusted.
- Capability inventory: Reviewing, writing, and refactoring source code.
- Sanitization: Absent; no guidance is provided to sanitize input code or escape generated content.
- [DYNAMIC_EXECUTION]: The rule file 'rules/rendering-hydration-no-flicker.md' provides a template for injecting an inline script via 'dangerouslySetInnerHTML' to prevent hydration flicker. While this is a standard React performance optimization, it involves the manual generation and execution of script content at runtime.
- [EXTERNAL_DOWNLOADS]: The skill recommends the installation and use of external third-party packages, specifically 'better-all' (github.com/shuding/better-all) and 'lru-cache' (github.com/isaacs/node-lru-cache), to achieve performance gains described in the guidelines.
Audit Metadata