llm-infrastructure

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • COMMAND_EXECUTION (SAFE): The skill includes utility scripts like fetch-openrouter-models.py and validate_llm_config.py. These scripts are transparent in their operation, performing legitimate tasks such as fetching model pricing from the OpenRouter API or scanning local code for deprecated model names and hardcoded keys. No malicious execution patterns were found.
  • EXTERNAL_DOWNLOADS (SAFE): The fetch-openrouter-models.py script makes network requests to openrouter.ai, a standard service for LLM access. The documentation mentions external packages like promptfoo and arize-phoenix, which are well-known tools in the AI observability and evaluation space.
  • CREDENTIALS_UNSAFE (SAFE): The provided scripts correctly use environment variables (e.g., OPENROUTER_API_KEY) and avoid hardcoding secrets. Furthermore, the validate_llm_config.py tool is specifically designed to detect and warn users about hardcoded credentials in their own projects, enhancing security posture.
  • DATA_EXPOSURE (SAFE): While validate_llm_config.py reads local source files, it does so to identify security anti-patterns and does not transmit the contents externally. The information fetched from OpenRouter is non-sensitive model metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 08:28 PM