The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses AI agents (Planner, Generator, Healer) to ingest data from external, potentially untrusted sources such as live web application snapshots, console logs, and network traces. * Ingestion points: Page snapshots and traces gathered by the @playwright/mcp server from target URLs. * Boundary markers: Not explicitly defined in the provided instruction set; relies on standard agent behavior. * Capability inventory: File writing (tests/.spec.ts, specs/.md, playwright.config.ts), shell command execution (npx playwright test, npm install), and browser control (browser_run_code). * Sanitization: The skill implements a 'Spec-first' flow where the user must review and approve the Markdown specification before code generation, and verifies generated tests using a test-provenance-guard.
[DYNAMIC_EXECUTION]: The skill's primary workflow involves generating executable TypeScript code (tests/*.spec.ts) and subsequently running it using the Playwright test runner. * Evidence: 'Step 2
Generate' produces test files, and 'Step 3
Run' executes them via npx playwright test. * Risk Mitigation: The skill prescribes a locator ladder and specific templates to restrict the range of generated code, and includes a verification phase with a provenance guard.
[COMMAND_EXECUTION]: The skill executes shell commands for project initialization, package installation, and test running. * Evidence: npx playwright init-agents, npm i -D @playwright/test@latest @playwright/mcp@latest, and npx playwright test. * Risk Mitigation: The skill includes a 'Phase 0
Preflight' gate that halts execution and prints the install plan for user approval before running any commands.
[EXTERNAL_DOWNLOADS]: The skill coordinates the installation of the @playwright/test and @playwright/mcp packages from the default npm registry. * Evidence: npm i -D @playwright/test@latest @playwright/mcp@latest in templates/install-plan.md.

e2e-testing