The Agent Skills Directory

Indirect Prompt Injection: The skill is designed to audit source code, which involves processing external data that could contain malicious instructions. While this is a common pattern for developer tools, it represents an attack surface where untrusted content enters the agent's context.
Ingestion points: Source code files (TypeScript/JavaScript) provided for memory leak auditing.
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded prompts within the code being audited.
Capability inventory: The skill primarily provides auditing logic but mentions verifying fixes using npm performance scripts.
Sanitization: No specific sanitization or filtering of the audited code is mentioned.
Command Execution: The instructions suggest running npm run perf:chat-leak to verify fixes. This is a standard project-specific development command used for performance regression testing and is used here in a benign, functional context.

memory-leak-audit