The Agent Skills Directory

Indirect Prompt Injection Surface: The skill establishes a feedback loop where user input entered into generated tools is converted into natural language instructions for the agent. This represents a potential indirect prompt injection surface as the agent processes content that originated from an external user interface without further sanitization.
Command Execution: The skill directs the agent to execute shell commands such as open to launch generated tools and git to populate review templates. These actions allow the agent to interact with the host system's file handler and version control utilities.
Dynamic Code Generation: The skill dynamically generates HTML and JavaScript files at runtime. While these files are intended for local user exploration and use pre-defined templates, the generation of executable content is a pattern that warrants awareness regarding the source and structure of the generated code.
Dependency Minimization: The skill explicitly requires that all generated playgrounds be self-contained with no external CSS or JavaScript dependencies. This is a security-positive practice that reduces exposure to external resource availability and potential supply chain risks.

playground