skills/aeondave/malskill/pytorch/Gen Agent Trust Hub

pytorch

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill contains only instructional content and code snippets for working with the PyTorch framework. No executable scripts, network requests, or hidden data exfiltration patterns were detected.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing external model checkpoints. Ingestion points: torch.load calls in SKILL.md. Boundary markers: The documentation instructs the user to use weights_only=True or safetensors. Capability inventory: The skill uses torch.load to deserialize data. Sanitization: The skill provides guidance on using add_safe_globals and warns against arbitrary code execution via pickle.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 10:39 PM
Security Audit — agent-trust-hub — pytorch