The Agent Skills Directory

[SAFE]: The skill functions as a text-generation assistant for drafting prompts. It does not utilize any capabilities for network communication, arbitrary code execution, or persistent access.
[SAFE]: The tool 'Read' is only used to access the provided markdown reference files (system-prompts.md and agentic-prompts.md), which contain educational content and best practices.
[SAFE]: User input provided via $ARGUMENTS is treated as a task description for the assistant. An evaluation of Category 8 (Indirect Prompt Injection) indicates a minimal risk surface: data enters via $ARGUMENTS in SKILL.md without boundary markers, but the agent's capability inventory is limited to local file reading and user interaction, with no path to sensitive data or privileged execution.

optimize-prompt