Galileo python sdk

• Outcome: Pass
• Latest audit: May 29, 2026, 8:38 PM
• Version:
  1.2.1

Overview

This is a documentation-only Galileo SDK reference, but its tracing examples can send LLM prompts, outputs, and related telemetry to Galileo if users enable them.

Install only if you intend to use Galileo observability or evaluation workflows. Before copying examples into real applications, confirm the packages are official, protect API keys, start with non-sensitive test data, and decide whether prompts, completions, retrieved documents, tool inputs, identifiers, and metadata are appropriate to send to Galileo. For production or regulated data, configure redaction, sampling, retention, access controls, or disable logging where needed.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• 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
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (6)

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill repeatedly promotes automatic tracing, observability, and runtime logging, but does not clearly warn that prompts, responses, and other conversational metadata may be transmitted to Galileo. In a security-sensitive agent context, this omission can cause developers to enable telemetry that exports sensitive user data, proprietary prompts, or regulated content without informed consent or data-handling review.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The wrapped OpenAI client example enables zero-config automatic tracing of all calls without warning that request and response bodies may be logged and exported. Because the example is copy-paste ready, users may unknowingly instrument live traffic and leak confidential prompts, retrieved context, or model outputs to an external telemetry service.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The flush instructions say to upload captured traces but do not clearly emphasize that calling flush transmits buffered trace contents to an external Galileo service endpoint. A developer may treat flush as a local cleanup step and accidentally send sensitive accumulated conversation or tool data off-system.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The documentation explicitly states that calls are automatically logged to Galileo with full span details, but it does not warn users that prompts, model inputs/outputs, metadata, and other potentially sensitive telemetry may be transmitted to an external service. In an observability SDK for GenAI systems, this omission is security-relevant because developers may enable tracing in production without understanding the data exposure implications.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The exporter example configures OpenTelemetry to send traces to a remote Galileo endpoint using an authorization header, but it does not include any warning that trace contents leave the local environment. Because GenAI traces often contain sensitive application and user data, omission of a disclosure and data-minimization guidance can lead to unintended external data transmission.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The documentation shows users how to authenticate and send evaluation traces, including inputs, outputs, and context, to Galileo, but it does not warn that these values may contain sensitive or regulated data that will be transmitted to an external service. In an LLM evaluation/RAG context, prompts and retrieved context frequently include proprietary, personal, or customer data, so omission of a disclosure materially increases the risk of accidental data exfiltration or policy violations.

VirusTotal

65/65 vendors flagged this skill as clean.