ClawHub
Back to skill

Security audit

Cognee Memory System

Security checks across malware telemetry and agentic risk

Overview

This is a coherent AI memory skill, but it needs Review because it describes automatic long-term capture of agent/session data without clear privacy, retention, or scoping controls.

Install only if you are comfortable with prompts, tool activity, and session context becoming persistent memory. Use scoped API keys, avoid secrets or regulated data, verify the external pip/npm packages, and confirm where memories are stored, who can access shared graphs, and how deletion or recovery works before enabling automatic hooks.

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 (2)

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill documents destructive deletion operations such as `cognee-cli forget --all` and `await cognee.forget(dataset="main_dataset")` without any warning, confirmation step, scope explanation, or recovery guidance. In a memory-management skill, this is materially risky because users or agents may execute the command directly and erase stored knowledge or session state unintentionally.

Missing User Warnings

High
Confidence
96% confidence
Finding
The plugin section describes automatic capture of actions and prompts and injection of relevant context across session lifecycle hooks, but provides no privacy notice, consent model, data minimization guidance, or retention boundaries. Because this skill is specifically a memory system that persists user and session information, silent collection and reinjection of potentially sensitive data can lead to privacy violations, over-collection, and unintended disclosure to later prompts or agents.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.