The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface inherent to its function as a web scraper.
Ingestion points: The skill uses page.goto to navigate to and process content from external URLs.
Boundary markers: None identified in the provided templates.
Capability inventory: The skill can execute browser-side logic via page.evaluate and perform interactions like page.click and page.type.
Sanitization: No sanitization is implemented for external content within the provided examples.
[DATA_EXFILTRATION]: The skill includes patterns for connecting to LambdaTest's cloud browser grid.
Evidence: Connection to wss://cdp.lambdatest.com/puppeteer for remote browser control.
Context: This is a standard connection to a well-known service for cross-browser testing.
[CREDENTIALS_UNSAFE]: The skill demonstrates safe credential handling for its cloud integration.
Evidence: Use of process.env.LT_USERNAME and process.env.LT_ACCESS_KEY.
Context: Storing sensitive keys in environment variables is a recommended security practice.

puppeteer-skill