The Agent Skills Directory

[SAFE]: The skill is a meta-utility designed to help users structure and optimize other agent skills. It follows best practices by encouraging the use of sibling files for long references and including evaluation hooks for testing.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests user-provided notes (notes.md) or prompts to generate new skill definitions. However, since the primary function is to assist in writing documentation and instructions, and it lacks dangerous execution capabilities or network exfiltration patterns, the risk is minimal.
Ingestion points: User prompts and external files like notes.md are used as inputs for generating skill content.
Boundary markers: The skill does not explicitly instruct the agent to ignore or delimit instructions found within the input notes.
Capability inventory: The skill is designed to produce file changes (creating/revising .md and .json files) based on user input.
Sanitization: There are no explicit sanitization or validation steps mentioned for the input content.

arc-creating-skill