tao-analyze-gaps-vlm-bcq
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized activities were detected. The skill is an official NVIDIA utility for analyzing model evaluation results within the TAO Toolkit environment.
- [COMMAND_EXECUTION]: The skill uses the
gap_analysiscommand to process local prediction files. This execution is performed within a designated container image (tao_toolkit.data_services) as defined in the configuration. - [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection due to its role in processing external VLM prediction data.
- Ingestion points: Data enters via the
predictions_jsonfile as described inSKILL.md. - Boundary markers: Absent. The skill does not provide markers to delineate untrusted content for downstream consumers.
- Capability inventory: The skill executes commands (
gap_analysis) and performs local file writes (SKILL.md). - Sanitization: Partial. The tool validates classification labels ('yes'/'no') but does not sanitize the full response text that is included in the generated reports.
Audit Metadata