The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes shell scripts in the scripts/ directory (check-server-actions-auth.sh, detect-barrel-imports.sh, and find-waterfall-chains.sh) that use standard Unix tools like find and grep to analyze local source code for performance and security issues. These scripts are used purely for diagnostic purposes and do not execute or modify the content they analyze.
[INDIRECT_PROMPT_INJECTION]: As a code-auditing skill, it has an inherent attack surface because it ingests and processes untrusted source code provided by the user. However, the risk is minimal as the skill's capabilities are restricted to identifying patterns and providing advice, without executing the untrusted data or transmitting it to external servers.
Ingestion points: Local source code files processed by automation scripts and the AI agent.
Boundary markers: None explicitly defined.
Capability inventory: Local file reading and pattern matching via grep and find; generation of architectural and security recommendations.
Sanitization: Standard LLM output guardrails are the primary defense.
[EXTERNAL_DOWNLOADS]: The documentation references external resources and well-known services (such as Vercel, Next.js, and specific GitHub repositories) for educational purposes. These are informational links and do not involve automated remote script execution or unverifiable package installations.

accelint-nextjs-best-practices