ClawHub
Back to skill

Security audit

Data Analyst

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only data analysis skill that tells the agent to work with user-uploaded files and shows no hidden code, persistence, credentials, or exfiltration behavior.

This skill is reasonable to install for analyzing files you intentionally upload. Treat uploaded datasets as visible to the agent session, and avoid uploading sensitive data unless you are comfortable with it being processed for the requested analysis.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Vague Triggers

Medium
Confidence
84% confidence
Finding
The skill is activated by very broad language such as 'Complete the data analysis tasks delegated by the user' without defining scope, trust boundaries, or restricted operations. In an agent setting, this can cause over-invocation and unsafe handling of untrusted data or files, increasing the chance the skill is used for actions beyond intended data analysis.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.