The Agent Skills Directory

[SAFE]: The skill consists entirely of Markdown-based instructions and reference materials. It does not perform any shell command execution, network requests, or file system modifications.
[DATA_EXPOSURE]: The instructions direct the agent to read specific local project files (e.g., .agents/product-marketing.md) to align edits with brand voice. This is a legitimate use of local context and does not involve accessing sensitive system credentials or private user data.
[PROMPT_INJECTION]: The skill processes user-supplied text for editing, which is an inherent surface for indirect prompt injection. However, the framework is purely focused on linguistic and stylistic analysis, and there are no instructions that would encourage the agent to execute commands or override safety boundaries based on the content being edited.

copy-editing