distributed-training-setup

Pass

Audited by Gen Agent Trust Hub on Feb 18, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [NO_CODE] (SAFE): No scripts, executables, or code blocks were found in the skill definition.
  • [COMMAND_EXECUTION] (SAFE): The skill metadata requests permissions for Bash (Python and Pip), but the skill does not actually implement any commands or logic to use these tools.
  • [INDIRECT_PROMPT_INJECTION] (LOW): The skill is designed to process user-provided requests for machine learning setup. Because the skill definition includes permissions for powerful tools like Bash(pip:*) and Write, it defines a potential attack surface for indirect prompt injection if a user or an external data source provides malicious instructions.
  • Evidence Chain (Category 8):
  • Ingestion point: User requests ('Help me with...').
  • Boundary markers: Absent.
  • Capability inventory: Bash(python:), Bash(pip:), Write, Edit.
  • Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 18, 2026, 03:03 PM