The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from various external websites, creating a potential surface for indirect prompt injection attacks.
Ingestion points: Web scraping results from Amazon, LinkedIn, YouTube, and other platforms, as well as arbitrary URLs via the scrape_url method.
Boundary markers: Absent. The instructions do not provide specific markers or guidance to separate scraped content from the agent's instructions.
Capability inventory: The skill includes network request capabilities and browser automation (CDP connection), allowing for follow-up actions based on scraped content.
Sanitization: Absent. There are no instructions for sanitizing or validating retrieved web content before the agent interprets it.
[COMMAND_EXECUTION]: The Browser API facilitates remote browser interaction via the Chrome DevTools Protocol (CDP), as detailed in references/advanced.md, allowing the agent to perform actions such as clicking, scrolling, and executing JavaScript within a remote browser session.
[EXTERNAL_DOWNLOADS]: The documentation references external automation frameworks including Playwright, Puppeteer, and Selenium, which are used to connect to the Bright Data Browser API for advanced web interactions.

brightdata-sdk