flask-werkzeug-attack
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides numerous bash commands using
curlto probe target web servers for debugger signatures and to trigger HTTP 500 errors for reconnaissance. - [REMOTE_CODE_EXECUTION]: The skill contains instructions and code snippets for exploiting the Werkzeug debugger console to execute arbitrary Python code. This behavior is documented as a method to attack a target server and does not target the agent's host environment.
- [DATA_EXFILTRATION]: Instructions are provided for extracting sensitive data such as server file paths, framework versions, and source code snippets from the traceback pages generated by a misconfigured Flask application.
- [SAFE]: The Python code provided for source code extraction uses standard libraries (
requests,re) and performs operations expected for its stated purpose of security reconnaissance.
Audit Metadata