The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to scrape content from external, untrusted websites, which can lead to indirect prompt injection if the agent processes the fetched content without proper sanitization.
Ingestion points: The stealthScrape method in index.js uses page.evaluate to extract text content from any user-provided URL.
Boundary markers: None. The scraped text is returned directly to the agent's context.
Capability inventory: The skill has capabilities for network access (page.goto), browser-side code execution (page.evaluate), file system access (fs.writeFile, fs.readFile), and visual monitoring (page.screenshot).
Sanitization: There is no evidence of content filtering, escaping, or validation of the data retrieved from external websites.
[COMMAND_EXECUTION]: The skill uses page.evaluate in index.js to execute JavaScript code within the Puppeteer browser environment. This is used to manipulate browser fingerprints and extract data from the DOM.
[DATA_EXFILTRATION]: The skill implements saveCookies and loadCookies methods in index.js that interact with the local file system using fs.writeFile and fs.readFile.
Evidence: These methods accept a filepath parameter from the caller, allowing the agent to read or write files at arbitrary locations. This capability could be abused to exfiltrate sensitive session data or overwrite critical system files if the agent is not restricted to a specific directory.

puppeteer-stealth