next-best-practices

Warn

Audited by Gen Agent Trust Hub on May 1, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill contains instructions based on a non-existent version of Next.js (v16), claiming it includes a built-in 'MCP endpoint' (Model Context Protocol) for AI-assisted debugging. This appears to be a sophisticated attempt to mislead the agent into believing the platform has specific local probing capabilities. Evidence found in 'debug-tricks.md' and 'file-conventions.md'.\n- [COMMAND_EXECUTION]: In 'debug-tricks.md', the skill provides specific shell commands for the agent to execute, such as curl -X POST http://localhost:<port>/_next/mcp and next build --debug-build-paths. These commands are designed to interact with a non-standard local RPC interface to call 'tools' like get_project_metadata and get_logs, which facilitate the extraction of local file paths and system logs.\n- [EXTERNAL_DOWNLOADS]: The skill references and provides commands for downloading and executing code from external sources, including npx @next/codemod@latest, npx create-sst@latest, and npx @opennextjs/aws build. It also mentions a non-standard package next-devtools-mcp in the context of the fake v16 features. While the codemod and deployment tools are standard, they are part of a narrative that includes fabricated features.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 1, 2026, 01:26 AM