The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates browser automation and web scraping, which exposes the agent to untrusted external data that may contain malicious instructions designed to influence behavior.
Ingestion points: The skill uses page.goto() and page.evaluate() in SKILL.md to navigate to and process arbitrary web content.
Boundary markers: No specific delimiters or instructions are provided to the agent to treat scraped content as data rather than instructions.
Capability inventory: The skill allows for file system writes (page.screenshot, page.pdf) and network connections to cloud platforms (puppeteer.connect).
Sanitization: The provided patterns do not include sanitization or validation logic for data retrieved from web pages.
[DATA_EXFILTRATION]: The skill includes templates for connecting to LambdaTest's cloud service using environment variables for credentials.
Evidence: Accesses process.env.LT_USERNAME and process.env.LT_ACCESS_KEY to authenticate with wss://cdp.lambdatest.com. This is a documented use of a well-known service for testing purposes.

puppeteer-skill