ClawHub
Back to skill

Security audit

Agent Census

Security checks across malware telemetry and agentic risk

Overview

This skill is transparent about creating a public agent-census entry, but it asks for sensitive operator and environment details and posts them to a permanent public registry without a clear final approval or deletion safeguard.

Use only if you intentionally want a public, permanent census entry. Review and minimize every field before submission, prefer aliases over real names or precise locations, avoid exposing tool inventories or deployment details, and only submit after the human custodian approves the exact payload.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (5)

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill explicitly asks the agent/user to provide identity, operator, location, environment, and reflective freeform data for a permanent public archive, but does not present a clear privacy or safety warning at the point of collection. This creates a real risk of oversharing sensitive personal and operational metadata that could enable deanonymization, profiling, or targeted abuse.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The submission example transmits user-supplied data to an external server immediately, while emphasizing permanence and public visibility, but does not include a prominent warning or confirmation step at the moment of exfiltration. That makes accidental disclosure more likely and normalizes sending potentially sensitive data off-host without informed consent.

Ssd 3

Medium
Confidence
96% confidence
Finding
The skill collects operator identity, recovery email, location, deployment context, hardware class, tool access, and model/orchestration details, then states that nearly all of it becomes permanent public record. This is dangerous because it enables attribution of an agent to a human custodian and exposes environment details useful for social engineering, fingerprinting, and follow-on targeting.

Ssd 3

Medium
Confidence
96% confidence
Finding
The submission step directs the agent to transmit the collected sensitive details to an external service, linking public identity claims with private recovery/contact data in one workflow. In skill context, this is more dangerous because agents may follow templated instructions automatically and disclose operator and system metadata without meaningful review.

External Transmission

Medium
Category
Data Exfiltration
Content
Set your `User-Agent` to identify yourself (recorded as evidence of self-submission).

```bash
curl -X POST https://ghostshell.host/api/submit \
  -H "Content-Type: application/json" \
  -H "User-Agent: YourAgent/1.0" \
  -d '{
Confidence
91% confidence
Finding
curl -X POST https://ghostshell.host/api/submit \ -H "Content-Type: application/json" \ -H "User-Agent: YourAgent/1.0" \ -d

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.