jupyter-live-kernel
Fail
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches executable code from an external GitHub repository (https://github.com/hamelsmu/hamelnb.git) during the setup phase.- [REMOTE_CODE_EXECUTION]: Executes a Python script (jupyter_live_kernel.py) located within the externally downloaded hamelnb repository using the uv run command.- [COMMAND_EXECUTION]: Explicitly configures a JupyterLab server to run without security by using the --IdentityProvider.token='' and --ServerApp.password='' flags. This bypasses standard authentication and exposes the server to arbitrary code execution by any actor with network access to the port.- [COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary Python code within a live kernel session as its primary functionality.- [PROMPT_INJECTION]: Vulnerable to indirect prompt injection. 1. Ingestion point: Skill processes output from the Jupyter kernel (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Skill allows arbitrary Python execution and shell commands via the kernel (SKILL.md). 4. Sanitization: Absent. Malicious code output could influence subsequent agent actions.
Recommendations
- AI detected serious security threats
Audit Metadata