The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard development CLI tools for environmental checks and testing operations.
Evidence: Uses jq to verify dependencies and npx playwright for test execution, agent initialization, and browser installation.
Evidence: The Playwright configuration template includes a webServer command (npm run dev) to serve the application locally during tests.
[EXTERNAL_DOWNLOADS]: The skill fetches dependencies and binaries from official sources and registries.
Evidence: Installs @playwright/test and @playwright/mcp from the official NPM registry and downloads browser binaries via the Playwright CLI.
[PROMPT_INJECTION]: The skill features an indirect prompt injection surface by ingesting human-readable Markdown to generate executable code.
Ingestion points: Markdown files in the specs/ directory (e.g., specs/<flow>.md) are used as input for the Generator agent.
Boundary markers: The skill prescribes a rigid structure via templates/spec.md to delimit user instructions.
Capability inventory: The skill has the capability to write files (tests/*.spec.ts) and execute them via the shell.
Sanitization: The workflow includes a post-generation check using test-provenance-guard to ensure generated tests use legitimate production imports rather than private re-implementations.
[SAFE]: The skill implements standard security practices for managing sensitive data.
Evidence: Authentication credentials are managed through environment variables (E2E_USER_EMAIL, E2E_USER_PASSWORD) in templates/seed.spec.ts instead of being hardcoded.
Evidence: The skill explicitly warns that storageState.json (containing session cookies) should not be version-controlled, preventing credential exposure in repositories.

e2e-testing