The Agent Skills Directory

[DATA_EXFILTRATION]: The skill uses the browser_screenshot tool to capture content from external URLs and saves the resulting files to the local filesystem (screenshots/ directory).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external web pages during the visual review phase, which could contain instructions designed to manipulate the agent's analysis.
Ingestion points: Data enters the system when the agent captures and reads screenshots of external websites (SKILL.md).
Boundary markers: There are no explicit markers or instructions to ignore embedded content within the analyzed images.
Capability inventory: The skill utilizes network access via browser_screenshot and file system access for reading and writing PNG files (SKILL.md).
Sanitization: No validation or sanitization of the visual content is specified before the visual review step.

visual-qa