The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables an agent to interact with and extract data from external websites using tools such as browser_snapshot, browser_network_requests, and browser_console_messages. This creates a surface for indirect prompt injection where malicious instructions embedded in a target website could influence the agent's behavior.
Ingestion points: Website semantic structure via browser_snapshot (SKILL.md), network traffic logs via browser_network_requests (SKILL.md), and browser console logs via browser_console_messages (SKILL.md).
Boundary markers: The skill does not provide instructions to the agent to treat retrieved website content as untrusted or to ignore instructions embedded within that content.
Capability inventory: Full browser control including browser_navigate, browser_click, browser_type, browser_evaluate, and browser_file_upload (SKILL.md).
Sanitization: There are no instructions for sanitizing or validating the external data before the agent processes it.
[DYNAMIC_EXECUTION]: The skill provides patterns and examples for using the browser_evaluate tool, which executes JavaScript functions in the browser context at runtime.
Evidence: Detailed documentation and examples for browser_evaluate in SKILL.md and examples/component-library.md.
[COMMAND_EXECUTION]: The skill facilitates complex interactions with a browser environment, including element clicking, text typing, and key presses, which constitute command execution within the browser's scope.
Evidence: Browser interaction tools like browser_click, browser_type, and browser_press_key are central to the skill's functionality as documented in SKILL.md.

playwright-visual-testing