The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection through its AI-powered testing lifecycle features, specifically the Planner, Generator, and Healer agents.
Ingestion points: AI agents ingest untrusted data from the application under test by processing its DOM, accessibility tree, and visible UI content during autonomous exploration and self-healing cycles (detailed in references/ai-powered-e2e-testing.md).
Boundary markers: The provided instructions do not include specific boundary markers or defensive prompts to instruct the AI agents to ignore potential instructions embedded within the application data being tested.
Capability inventory: The skill leverages capabilities that could be exploited if an agent were subverted, such as executing shell commands via execSync for environment setup (references/environment-management.md), writing files for test generation, and performing network operations for API validation and cloud device farm interaction (references/api-e2e-testing.md and references/cloud-testing.md).
Sanitization: There are no documented sanitization or validation mechanisms for content retrieved from external applications before it is utilized by the AI testing agents.

voyager