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_analysis command 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_json file as described in SKILL.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
Risk Level
SAFE
Analyzed
Jun 30, 2026, 05:48 PM
Security Audit — agent-trust-hub — tao-analyze-gaps-vlm-bcq