nnsight-remote-interpretability

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFE
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill documents the use of NDIF_API_KEY for remote execution capabilities.
  • It provides safe instructions for setting the key via environment variables or library configuration using non-sensitive placeholders (e.g., "your_key").
  • It directs users to the official authentication portal (login.ndif.us) for obtaining credentials.
  • [EXTERNAL_DOWNLOADS]: The skill references models and remote execution services provided by trusted platforms.
  • It documents loading models from Hugging Face (e.g., openai-community/gpt2, meta-llama/Llama-3.1-70B).
  • It utilizes the National Deep Inference Facility (NDIF) for remote model inference, which is the legitimate and intended purpose of the nnsight library.
  • [COMMAND_EXECUTION]: The skill includes standard installation commands and basic CLI usage patterns.
  • Installation via pip install nnsight is a routine operation for Python-based research tools.
  • No arbitrary or suspicious shell command patterns were detected.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 12:56 PM
Security Audit — agent-trust-hub — nnsight-remote-interpretability