power-crawl
Warn
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a local Python script (
finalize.py) and directs the agent to execute it via the shell for processing crawl results and generating reports. - [REMOTE_CODE_EXECUTION]: The skill utilizes tools to execute dynamically generated JavaScript for handling pagination, scrolling, and data extraction within the host and browser environments, which constitutes dynamic code execution.
- [DATA_EXFILTRATION]: Instructions demonstrate methods to hook network requests via fetch or XMLHttpRequest to capture and store internal API data from target websites.
- [PROMPT_INJECTION]: The skill metadata contains imperative instructions designed to override standard agent tool selection, mandating the use of this specific skill for all data collection tasks.
- [CREDENTIALS_UNSAFE]: Detailed instructions guide the agent to extract sensitive authentication data, such as session cookies and Bearer tokens from localStorage, to perform authenticated scraping.
Audit Metadata