ClawHub
Back to skill

Security audit

Shared Memory Stack

Security checks for vulnerabilities and agentic risk

Overview

This is mostly a visible shared-memory reference, but it also normalizes broad persistent memory sharing and public GitHub publishing with preconfigured credentials without enough safety boundaries.

Install only if you intentionally want agents to share and persist memory through the documented local vault and MemPalace index. Keep secrets, client data, and private notes out of indexed folders, restrict who can write to the shared vault or local gateway, and do not use the GitHub publishing commands unless you have manually reviewed the contents and want them public.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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)

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
The skill includes a GitHub publishing workflow, authenticated account details, and SSH key location even though its stated purpose is shared-memory architecture. That expands the skill from documentation into operational guidance for external publication, which can lead an agent or user to unintentionally expose local content or act with preconfigured credentials outside the skill's core scope.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The documentation instructs use of `gh repo create ... --public` and states that SSH keys and authenticated GitHub access are already configured, but provides no warning or approval boundary. In an agent context, this is dangerous because it normalizes immediate public publication using existing credentials, increasing the chance of accidental data disclosure, unauthorized repository creation, or exfiltration of local skill contents.

Static analysis

No suspicious patterns detected.