python-development
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for ingesting untrusted data through network requests, such as the
httpxexample inSKILL.md. Mandatory evidence chain: 1. Ingestion points:fetch_datafunction. 2. Boundary markers: Absent in provided examples. 3. Capability inventory:Bash,Write,Edit, andNotebookEdittools are permitted. 4. Sanitization: Absent in the documentation. This combination poses a risk if remote content is processed without adequate validation. - [COMMAND_EXECUTION]: The skill provides instructions for executing system commands for profiling and testing using
uv runandBash. - [EXTERNAL_DOWNLOADS]: The skill documents configuration and dependencies from well-known sources, including tools from Astral, PyCQA, and Codecov.
- [DYNAMIC_EXECUTION]:
REFERENCE.mdcontains an instruction to disable the security PIN for the Werkzeug debugger (os.environ['WERKZEUG_DEBUG_PIN'] = 'off'). While aimed at development convenience, this practice removes the protection that prevents unauthorized remote code execution via the debugger's web interface.
Audit Metadata