The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates interaction with external web content, which creates a surface for indirect prompt injection.
Ingestion points: The agent context ingests untrusted data when the browser navigates to or verifies content on web pages (e.g., to LoginPage, at DashboardPage, welcomeMessage.text()) as seen in SKILL.md and reference/playbook.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded within the processed web content.
Capability inventory: The skill possesses capabilities including network access via RemoteWebDriver, dynamic browser-based execution via js.exec, and file system access for test reporting and uploads.
Sanitization: No evidence was found of content filtering or sanitization of text retrieved from web elements before it is used in logic or assertions.
[EXTERNAL_DOWNLOADS]: The skill configuration in reference/playbook.md references external dependencies downloaded via Gradle from Maven Central.
It fetches established testing libraries such as org.gebish:geb-spock, org.seleniumhq.selenium:selenium-java, and io.github.bonigarcia:webdrivermanager.
[DYNAMIC_EXECUTION]: The skill instructions in reference/playbook.md include patterns for executing JavaScript within the browser context using js.exec.
This functionality is used for standard automation tasks like scrolling elements into view or retrieving page metadata.
[DATA_EXFILTRATION]: An automated scan flagged the URL https://hub.lambdatest.com/wd/hub as a phishing risk.
This endpoint is the official hub for LambdaTest cloud execution, which is the primary service provided by the skill author, lambdatest. It is used to transmit test commands and capabilities to the vendor's cloud infrastructure.

geb-skill