The Agent Skills Directory

Data Exposure (LOW): The skill is designed to automatically read local configuration files like package.json, tsconfig.json, and webpack/vite configurations, as well as arbitrary source files derived from stack traces. While this is necessary for its debugging purpose, it creates a surface for accessing sensitive project metadata if the agent is not restricted to the project root.
Indirect Prompt Injection (LOW): The skill processes untrusted data which could contain malicious instructions. Ingestion points: Console logs, error messages, and the contents of source files targeted for debugging. Boundary markers: None provided to separate input data from instructions. Capability inventory: Local file system read access and code generation. Sanitization: No evidence of input filtering or escaping. An attacker could place instructions inside code comments or log files to influence the agent's behavior during a debugging session.

smart-debugger