Gemini Deep Research

• Outcome: Pass
• Latest audit: May 29, 2026, 8:40 PM
• Version:
  2.0.0

Overview

This is a straightforward Gemini deep-research wrapper that sends research requests to Google and saves reports locally, with no hidden or destructive behavior found.

Install only if you are comfortable sending research prompts, formatting instructions, and any selected Gemini file-search-store context to Google. Prefer using GEMINI_API_KEY from the environment rather than passing keys on the command line, and choose a safe output directory or delete generated markdown and JSON files when research topics are sensitive.

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 (4)

Missing User Warnings

Low

• Confidence: 93% confidence
• Finding: The skill explicitly states that it writes a final report and full interaction metadata to timestamped files, but it does not warn users that content derived from prompts, web research, and model interactions will be persisted to disk. This can expose sensitive research topics, collected data, or API-related metadata to other local users, backups, or source control if the output directory is not handled carefully.

Missing User Warnings

Low

• Confidence: 91% confidence
• Finding: The skill describes systematic web searching and requires a GEMINI_API_KEY, but it does not disclose the privacy implications of sending user research topics to external services and performing repeated web queries. Users may unintentionally submit confidential topics, internal plans, or regulated data to third-party APIs and search targets without informed consent.

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The script sends the user-supplied research query, and optionally references a file search store, to an external Google API without an explicit consent prompt or prominent disclosure at the point of transmission. In a skill/agent context, users may supply sensitive business, personal, or proprietary data, making silent transmission to a third party a real privacy and data-governance risk.

Missing User Warnings

Low

• Confidence: 80% confidence
• Finding: The script writes both the generated report and the full JSON response to disk by default, potentially persisting sensitive prompts, outputs, metadata, and other returned content without a strong warning. In shared environments or agent workflows, this can create unintended local data exposure and retention risk.

VirusTotal

65/65 vendors flagged this skill as clean.