The Agent Skills Directory

[SAFE]: The skill is a meta-instruction set for generating other skills. It does not contain executable scripts, network requests, or operations that access sensitive system locations.
[PROMPT_INJECTION]: The instructions include guidance on how to write descriptions to ensure the agent correctly triggers a skill (e.g., 'Make sure to use this skill whenever...'). This is legitimate prompt engineering intended to improve model routing and does not attempt to bypass safety filters or ignore system-level constraints.
[INDIRECT_PROMPT_INJECTION]: The workflow involves taking user requirements to generate file content. This creates an interface for processing external data. The skill mitigates this through a structured 'Review with user' phase, allowing for manual verification of the generated content before it is finalized. The risk is considered low and aligned with the skill's primary function as a content generator.

create-skill