The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands like open, start, and xdg-open to launch files. On macOS, it uses osascript (AppleScript) to programmatically activate the browser and manage tabs, which represents automated UI control. It also retrieves the active tab's URL to verify the render.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data to generate HTML files that are automatically opened in the browser, creating an attack surface for Cross-Site Scripting (XSS).
Ingestion points: The skill ingests user-provided plans, layouts, and design descriptions in SKILL.md.
Boundary markers: Instructions require the agent to create "complete, self-contained HTML" and explicitly forbid "external requests" or "CDN links" in SKILL.md.
Capability inventory: Writing local files, executing shell commands (osascript, open), and controlling browser tabs (SKILL.md).
Sanitization: The skill mandates inlining all CSS and assets while disabling network-dependent features in the output (SKILL.md).
[DYNAMIC_EXECUTION]: The skill dynamically generates and executes AppleScript code blocks to control the state of Google Chrome windows and tabs based on file paths at runtime.

show-me