ClawHub

Vector Memory Hack

Security checks across malware telemetry and agentic risk

Overview

This is a local memory-search helper, but it is too broad about when to search and explicitly encourages surfacing credentials from memory.

Install only if you are comfortable with the agent indexing local memory into SQLite. Avoid storing credentials in MEMORY.md, use this skill only when the current task truly needs saved context, and delete or rebuild the index after removing sensitive entries.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Lp3

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding
The skill clearly directs the agent to read local memory/documentation files and exposes configurable paths for MEMORY.md and a SQLite database, but it declares no corresponding permissions or trust boundary. Undeclared file-read capability is dangerous because it can cause agents or platforms to grant broader filesystem access implicitly, reducing reviewability and increasing the chance of unauthorized access to sensitive local data.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The integration guidance explicitly tells agents to search memory for 'server addresses and credentials' before executing tasks, normalizing retrieval of highly sensitive secrets without consent, minimization, or handling safeguards. In this context, the skill is more dangerous because it is designed as a routine pre-task step, which can institutionalize unnecessary secret access and increase the chance of downstream leakage into logs, prompts, tool outputs, or actions.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal