The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides patterns for processing web content via Selenium, which introduces a surface for indirect prompt injection. This is effectively mitigated through provided best practices.
Ingestion points: Data is ingested via the driver.getPageSource() method in the BaseTest.java example found in references/page_object_model.md.
Boundary markers: The skill includes a dedicated 'Security Considerations' section in SKILL.md advising users to only test their own applications and avoid navigating to third-party sites.
Capability inventory: The skill includes browser automation capabilities (driver.get) and filesystem writes via a scaffolding script.
Sanitization: The provided implementation in BaseTest.java includes a 50,000 character truncation limit for page source attachments to prevent excessively large or malicious payloads from entering the AI context.
[COMMAND_EXECUTION]: The PowerShell script scripts/setup-maven-project.ps1 is used to scaffold new Maven projects. It performs standard filesystem operations (New-Item, Set-Content) using hardcoded local templates and does not execute untrusted remote code.
[EXTERNAL_DOWNLOADS]: The Maven project templates (scripts/pom-template.xml) utilize well-known, versioned dependencies from trusted organizations (SeleniumHQ, JUnit, AssertJ). These packages are resolved through official registries during the standard build process.
[SAFE]: The skill includes demo credentials (e.g., 'standard_user', 'secret_sauce') in its testing examples. These are standard placeholders for public testing sites and do not represent a security risk.

webapp-selenium-testing