Azure Ai Evaluation Py

• Outcome: Pass
• Latest audit: May 27, 2026, 9:48 PM
• Version:
  0.1.0

Overview

This skill is a coherent Azure AI evaluation helper; it uses Azure credentials and may send evaluation data to Azure services, but that behavior is disclosed and aligned with its purpose.

Install only if you intend to use Azure AI Evaluation. Run it in a virtual environment, verify the PyPI package source, use least-privilege Azure credentials, and review/redact evaluation datasets before using AI-assisted evaluators or Foundry logging because prompts, responses, context, ground truth, metrics, and result rows may be sent to or stored in Azure services.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

• Confidence: 78% confidence
• Finding: The trigger list contains broad phrases such as 'evaluate' and 'evaluators' that may activate this skill for unrelated requests. Over-broad activation can cause accidental invocation of functionality that reads environment configuration, processes local files, or sends evaluation data to external services in contexts where the user did not intend to use this skill.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The skill includes an example that logs evaluation results to a remote Foundry project without an explicit warning that prompts, responses, contexts, and derived metrics may be transmitted off-box. In an evaluation workflow, those artifacts can contain sensitive user inputs, proprietary documents, or model outputs, so silent remote logging creates a real data-exfiltration and compliance risk.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The prompt-based evaluator examples send query, response, and optional context to a remote Azure OpenAI service, but the documentation does not warn users that potentially sensitive application data may leave the local environment. In an evaluation SDK context, these fields often contain user prompts, model outputs, grounding context, or proprietary data, so omission of a clear disclosure can lead to unintended data exposure and compliance issues.

VirusTotal

65/65 vendors flagged this skill as clean.