The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to implement a postinstall.js script that programmatically modifies files within the node_modules directory. This pattern of 'patching' framework internals is highly brittle and can be used to hide malicious code modifications.
[REMOTE_CODE_EXECUTION]: Multiple instructions involve the use of npx to execute remote codemods (@next/codemod@canary) and shell-based build commands. While these tools are standard in the ecosystem, the skill's reliance on them to fix 'Critical' errors in a non-existent framework version increases the risk of executing untrusted logic.
[EXTERNAL_DOWNLOADS]: The skill recommends installing specific, unverified versions of the Next.js package (e.g., next@16.0.10) to mitigate a cited (but non-existent) RCE vulnerability, which could lead to the installation of compromised or malicious packages.
[DATA_EXPOSURE]: Documentation for the build command in references/patches.md suggests including environment variables like BETTER_AUTH_SECRET directly in the command line, which can lead to secret exposure in CI/CD logs or shell history.

nextjs