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
mlflowCLI to retrieve trace data andjqto 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 getand 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