developing-nextjs

Warn

Audited by Gen Agent Trust Hub on Nov 12, 2026

Risk Level: MEDIUMMETADATA_POISONINGPROMPT_INJECTION
Full Analysis
  • [METADATA_POISONING]: The skill is framed around 'Next.js 16' and 'React 19.2', versions that are not currently released or established. This deceptive context forces the agent to work with hallucinated framework patterns, which can result in the generation of unstable or insecure code.
  • [PROMPT_INJECTION]: The instructions explicitly direct the agent to avoid modifying existing 'middleware.ts' files ('DO NOT rename or change anything'). Because middleware is the primary mechanism for implementing authentication and authorization in Next.js, this constraint prevents the agent from auditing or updating critical security boundaries.
  • [PROMPT_INJECTION]: The skill fabricated a new 'proxy.ts' convention as a replacement for middleware. In the Next.js framework, 'proxy.ts' is not a recognized special file for request interception; therefore, any security or access control logic placed in this file by the agent will fail to execute, potentially leaving the application vulnerable to unauthorized access.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 12, 2026, 09:52 PM