The Agent Skills Directory

[DATA_EXFILTRATION]: Audits local repository configuration and source files, such as package.json, tailwind.config.*, and globals.css, to extract design tokens. The information gathering is strictly local for documentation purposes, and no network exfiltration patterns were observed.
[COMMAND_EXECUTION]: Instructs the agent to create and update documentation files and a new agent skill file (.agents/skills/frontend-design/SKILL.md) using local templates. These actions represent standard file system management for repository setup.
[PROMPT_INJECTION]: The skill audits pre-existing repository documentation and source code, which provides an indirect prompt injection surface. However, the skill primarily summarizes this content into a new document and requires user approval for any new design proposals, limiting the impact of potentially malicious content in the source files.

design-md