ClawHub
Back to skill

Security audit

EternalClaw Memory Sync

Security checks across malware telemetry and agentic risk

Overview

This skill is not clearly malicious, but it restores downloaded memory files directly into a workspace with weak scoping and overwrite protections.

Install only if you trust the publisher and the encrypted backup source. Restore into a new temporary directory first, inspect the resulting files, back up anything that could be overwritten, and avoid passing sensitive passwords directly on the command line when possible.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding
The skill documentation advertises network access, downloading remote content, and restoring files into the workspace, yet it declares no permissions. This creates a trust and review gap: operators may approve or run the skill without understanding that it can read/write local files and fetch attacker-controlled remote data.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The documented purpose claims secure backup/restore, but the described behavior includes writing restored contents to local files/directories under a chosen path and does not implement the claimed backup flow. This mismatch is dangerous because it obscures the real attack surface: a remote blob can drive workspace modification, potentially overwriting files or planting content, while users may trust it as a narrowly scoped sync utility.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The documentation instructs users that restored memory files will be unpacked into the current workspace without warning about overwrite, file planting, or broader workspace modification risks. In the context of downloading from a remote URL, this makes the skill more dangerous because users are encouraged to import external content directly into a live agent environment.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
This script downloads attacker-controlled content from a URL, decrypts it, and writes arbitrary file contents to disk with no user confirmation, dry-run mode, or restriction to a dedicated restore directory. Even though there is a basic path check, the operation can still overwrite files within the chosen output tree and materially modify the local filesystem, which is risky for a restore tool handling remote input.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.