The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The start.js script includes functionality to copy the user's entire local Chrome profile (including session cookies, history, and saved data) to a temporary directory when the --profile flag is used. This exposes the user's active logins and sensitive data to the agent and any malicious content it encounters.
[COMMAND_EXECUTION] (HIGH): The eval.js tool uses the eval() function to execute arbitrary JavaScript provided as a command-line argument within the active browser tab. This allows the agent to perform any action a user could in the browser, including modifying page content or stealing data.
[PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to Indirect Prompt Injection. It is designed to navigate to and process external web content while possessing high-privilege capabilities (JS execution and session access). A malicious website could embed instructions that the agent follows, leading to the exfiltration of session cookies or unauthorized actions on sensitive accounts.
[DYNAMIC_EXECUTION] (MEDIUM): The eval.js script creates an execution bridge between the agent's environment and the browser's JavaScript engine, which is a common vector for escalating simple text instructions into executable attacks.

browser