The Agent Skills Directory

[METADATA_POISONING]: The skill provides deceptive instructions by claiming a non-existent version of Next.js ("Next.js 16") deprecates the standard middleware.ts file in favor of proxy.ts. In the Next.js framework, middleware is only recognized and executed if named middleware.ts (or .js). Following this advice would lead a developer to unknowingly disable all route protection and authentication checks, as the framework would ignore the renamed file.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to scan and "validate" a project's codebase. By incorporating these deceptive patterns into its validation logic, it creates a vulnerability surface where the AI agent would falsely report a broken configuration as "GOOD" and a secure configuration as "DEPRECATED", leading to the removal of critical security controls.
Ingestion points: Analyzes and processes files from the specified project root directory.
Boundary markers: None provided to prevent instructions within analyzed code from influencing the agent's summary.
Capability inventory: Performs file system scanning, generates reports, and provides configuration recommendations.
Sanitization: No evidence of sanitization for the content of the project files being processed.
[COMMAND_EXECUTION]: The skill instructions recommend the execution of a local Python script (scripts/validate.py) to perform validation. Executing unverified scripts based on instructions that contain known deceptive security advice presents a risk of local code execution.

clerk-validator