ClawHub
Back to skill

Security audit

LYRA 3-Brain Memory

Security checks for vulnerabilities and agentic risk

Overview

This skill provides local persistent memory tools with explicit consent gates and disclosed disk-writing behavior.

Install only if you want a local, persistent memory journal for LYRA. Set LYRA_CORE_ROOT or LYRA_CORE to a directory you control, use write commands only when you intend disk persistence, and avoid storing secrets or sensitive personal data because written memory remains until you delete it.

SkillSpector

Vulnerability Patterns
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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)

Tp4

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding
The skill documentation overstates safety guarantees and understates actual behavior. In particular, saying writes occur under `LYRA_CORE_ROOT/memory` while also using `LYRA_CORE`, and requiring users to 'read references/SECURITY.md first' without enforcement, can mislead an agent or operator about where data may be written and what safeguards are actually active. In a persistence skill, documentation drift matters because it can cause unintended disk writes, unsafe path selection, or premature trust in security controls that do not exist.

Static analysis

No suspicious patterns detected.