The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for scraping and processing untrusted data from external websites, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through methods like page.content(), textContent(), and allTextContents() in references/scraping/extraction.md, as well as through intercepted network responses in references/scraping/network.md.
Boundary markers: The provided code examples do not include explicit delimiters or "ignore embedded instructions" warnings for the agent when processing scraped content.
Capability inventory: The skill enables network operations for scraping, file system writes via fs.writeFile and download.saveAs (documented in references/scraping/extraction.md and references/scraping/network.md), and browser-side script execution via page.evaluate().
Sanitization: The extraction logic demonstrated in the skill does not implement sanitization, escaping, or validation of the retrieved external content before it is processed.
[EXTERNAL_DOWNLOADS]: The skill references standard dependencies and configuration steps from well-known sources.
Uses official GitHub Actions (actions/checkout, actions/setup-node, actions/cache, actions/upload-artifact) for CI/CD workflows in references/e2e/ci.md.
Recommends installation of the playwright and @playwright/test packages from the standard NPM registry.
Suggests using ws for mock server testing and p-limit for concurrency management, which are common community packages.

playwright