The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the generation of test suites by ingesting source code files, which creates an attack surface for indirect prompt injection if the source code contains adversarial instructions designed to influence the agent.
Ingestion points: Processes source files (e.g., .ts, .tsx, .vue) to generate tests.
Boundary markers: No explicit delimiters or instructions to treat source content as untrusted are provided.
Capability inventory: The skill is designed to write test files and execute them through shell-based test runners (Jest, Vitest, Cypress).
Sanitization: No validation or sanitization of input source code is specified.
[CREDENTIALS_UNSAFE]: Multiple reference files include hardcoded credentials within testing examples.
Evidence: references/cypress-e2e-testing.md uses "user@example.com" and "password123" in E2E tests. references/react-testing-library-integration-tests.md uses "john@example.com" and "password123" in integration tests.
Context: While these are intended as documentation placeholders for local development (localhost), they present a security risk if used as templates in production-facing tests.

frontend-testing