Deep Researcher

• Outcome: Pass
• Latest audit: May 30, 2026, 9:13 PM
• Version:
  1.1.0

Overview

This is a research-writing skill that uses expected web and academic-source lookups, with no evidence of hidden, destructive, or credential-stealing behavior.

Install if you want an agent to perform broad web-based academic research and write long papers. Do not use confidential project names, unpublished research, personal data, or regulated information as topics unless you are comfortable with derived search queries being sent to third-party sites and public APIs.

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 (7)

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The trigger list contains broad phrases like 'academic paper', 'literature review', and 'comprehensive analysis' that can match many ordinary user requests. Overbroad activation can cause the skill to invoke unexpectedly, leading to unnecessary external searches, excessive data collection, or unintended delegation of user tasks to a high-privilege research workflow.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill prominently instructs use of broad external web research and multiple third-party data sources, but it does not clearly warn users that their prompts may be transmitted to outside services and websites. This creates a privacy and data-governance risk, especially if users include sensitive research topics, unpublished material, or regulated information in their requests.

External Transmission

Medium

• Category: Data Exfiltration

Content

- **arXiv**: `https://export.arxiv.org/api/query` — cutting-edge AI/ML/theory, no API key
- **Google Scholar**: `batch_web_search` — comprehensive peer-reviewed coverage
- **PubMed/PMC**: `https://eutils.ncbi.nlm.nih.gov/entrez/eutils/` — biomedical, life sciences
- **Semantic Scholar**: `https://api.semanticscholar.org/` — CS academic sources

### Economic & Policy
- **World Bank**: `https://api.worldbank.org/v2/` — free, no key required

• Confidence: 88% confidence
• Finding: https://api.semanticscholar.org/

External Transmission

Medium

• Category: Data Exfiltration

Content

- **Semantic Scholar**: `https://api.semanticscholar.org/` — CS academic sources

### Economic & Policy
- **World Bank**: `https://api.worldbank.org/v2/` — free, no key required
- **IMF**: `https://api.imf.org/` — macroeconomic data
- **OECD**: `https://stats.oecd.org/` — comparative policy data

• Confidence: 88% confidence
• Finding: https://api.worldbank.org/

External Transmission

Medium

• Category: Data Exfiltration

Content

### Economic & Policy
- **World Bank**: `https://api.worldbank.org/v2/` — free, no key required
- **IMF**: `https://api.imf.org/` — macroeconomic data
- **OECD**: `https://stats.oecd.org/` — comparative policy data

### Technology & AI

• Confidence: 88% confidence
• Finding: https://api.imf.org/

External Transmission

Medium

• Category: Data Exfiltration

Content

- **OECD**: `https://stats.oecd.org/` — comparative policy data

### Technology & AI
- **Hugging Face**: `https://api.huggingface.co/` — ML models, datasets, papers
- **GitHub API**: `https://api.github.com/` — code trends, repositories
- **Google Patents**: `https://patents.google.com/` — innovation trends

• Confidence: 89% confidence
• Finding: https://api.huggingface.co/

External Transmission

Medium

• Category: Data Exfiltration

Content

### Technology & AI
- **Hugging Face**: `https://api.huggingface.co/` — ML models, datasets, papers
- **GitHub API**: `https://api.github.com/` — code trends, repositories
- **Google Patents**: `https://patents.google.com/` — innovation trends

### Industry

• Confidence: 89% confidence
• Finding: https://api.github.com/

VirusTotal

66/66 vendors flagged this skill as clean.