The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and extract information from external websites using tools like browser_act and browser_get_page_info, which creates a vulnerability to malicious instructions embedded in web content.
Ingestion points: Data is ingested from any URL provided to browser_act, browser_get_page_info, or browser_manage_tabs within SKILL.md.
Boundary markers: The instructions lack explicit directives for the agent to distinguish between its primary instructions and potential instructions found within the scraped web content.
Capability inventory: The skill provides capabilities to navigate the web, extract text/tables, and perform UI actions like typing or clicking, which could be abused if the agent follows instructions found on a malicious page.
Sanitization: There is no evidence of sanitization or filtering of the retrieved web content before it is processed by the agent's language model.

browser-automation