analyzing-mlflow-trace

Pass

Audited by Gen Agent Trust Hub on Aug 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the mlflow CLI to retrieve trace data and jq to parse the resulting JSON files. These are legitimate, standard tools for MLflow trace analysis.
  • [PROMPT_INJECTION]: The skill processes MLflow traces, which are generated by external AI/ML applications. This creates an indirect prompt injection surface where malicious content within the traced application's inputs or outputs could attempt to influence the agent performing the analysis.
  • Ingestion points: Trace JSON data fetched via mlflow traces get and stored in /tmp/trace.json.
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions for the content within trace attributes.
  • Capability inventory: The skill uses shell command execution (mlflow, jq) and file system writes (/tmp/).
  • Sanitization: The skill does not prescribe sanitization or validation of the span attribute content before the agent evaluates it.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 11, 2026, 02:51 PM
Security Audit — agent-trust-hub — analyzing-mlflow-trace