ClawHub
Back to skill

Security audit

Academic Deep Research Pro

Security checks across malware telemetry and agentic risk

Overview

This is a markdown-only academic research workflow that uses web research and optional sub-agents in a disclosed, purpose-aligned way.

Before installing, understand that this skill is not truly offline: it is designed to search and fetch web content, and it may delegate independent research themes to sub-agents. Review the proposed research plan carefully, set source and scope limits, and avoid approving research that would expose confidential, regulated, or sensitive internal information.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Vague Triggers

Medium
Confidence
88% confidence
Finding
The invocation guidance allows triggering the skill by merely mentioning broad natural-language phrases like "deep research" or "exhaustive analysis," which can easily appear in ordinary conversation. This raises the risk of unintended activation, causing the agent to enter a tool-using research workflow without explicit user intent, potentially leading to unnecessary browsing, data exposure, or costly actions.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
---

## Phase 3: Research Cycles (Auto-Execute)

### Theme 1: Market Landscape — Cycle 1
Confidence
80% confidence
Finding
The explicit 'Auto-Execute' phase indicates the workflow proceeds through multiple research cycles without an intervening approval gate. In an agentic system, autonomous execution can amplify mistakes, increase exposure to malicious web content, and cause excessive or unintended external actions before the user can review or constrain scope.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.