The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted content from external websites via playwright-simple.js and playwright-stealth.js. This creates a surface for indirect prompt injection attacks.
Ingestion points: The scripts extract text using document.body.innerText and the full page source using document.documentElement.outerHTML from any provided URL.
Boundary markers: Scraped data is encapsulated in a JSON structure, providing a degree of separation, but the content values are returned raw.
Capability inventory: The skill has network access to fetch web content and file system access to save screenshots and HTML files (e.g., using fs.writeFileSync).
Sanitization: No sanitization or filtering of the extracted web content is performed before it is provided to the agent.
[DYNAMIC_EXECUTION]: The scraping logic utilizes page.evaluate() to execute JavaScript within the browser context of the target website. This is an inherent part of dynamic web scraping but involves executing code against untrusted external content.
[EXTERNAL_DOWNLOADS]: The skill documentation and package.json recommend installing the playwright package and its browser binaries via npm install and npx playwright install chromium. These resources are fetched from official package registries and well-known vendor sources.

playwright-scraper-skill