analyzing-mlflow-session

Pass

Audited by Gen Agent Trust Hub on Aug 11, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides documentation and patterns for using the mlflow CLI to debug multi-turn chat sessions.
  • [COMMAND_EXECUTION]: The skill uses mlflow traces, jq, cat, and head. These are standard tools for interacting with MLflow and processing JSON data. All shell examples use single quotes to prevent command injection via backticks in filters.
  • [DATA_EXFILTRATION]: The skill documents how to extract fields from traces but does not contain patterns for exfiltrating data to external servers. It correctly recommends writing large outputs to local temporary files (/tmp/) to avoid shell buffer limits.
  • [REMOTE_CODE_EXECUTION]: No remote code execution patterns were found. The referenced bash scripts in the scripts/ directory are described as local tools for schema discovery and turn inspection.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 11, 2026, 02:51 PM
Security Audit — agent-trust-hub — analyzing-mlflow-session