The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with untrusted external web content, which serves as a potential vector for indirect prompt injection.
Ingestion points: External data is ingested from web pages through tools such as navigate_page, take_snapshot, list_console_messages, and list_network_requests (SKILL.md).
Boundary markers: There are no explicit instructions or delimiters defined to separate untrusted web content from the agent's core instructions.
Capability inventory: The agent has high-privilege interaction capabilities including evaluate_script (JS execution), fill_form, and upload_file (SKILL.md).
Sanitization: No specific sanitization or filtering logic for processed web data is mentioned.
[DYNAMIC_EXECUTION]: The skill includes the evaluate_script tool, which enables the dynamic execution of JavaScript code within the context of a browser page for debugging and inspection purposes.
[COMMAND_EXECUTION]: The skill provides a comprehensive suite of browser control commands, including navigation (navigate_page), element interaction (click, hover, drag), and page inspection (take_screenshot, take_snapshot).

chrome-devtools