ClawHub
Back to skill

Security audit

Multi-AI Local Memory｜多 AI 本地共享記憶

Security checks for vulnerabilities and agentic risk

Overview

The skill’s shared-memory behavior matches its stated purpose, with privacy considerations and a separately installed runtime to review.

Local storage does not guarantee local-only processing: recalled information may reach connected AI providers. Connect only providers you trust with that information, keep secrets out, and review the separate runtime’s permissions, deletion controls, and sharing settings before installing.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

Automatic memory injection lacks a warning that recalled local data may be transmitted to external AI providers.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The skill requires automatic recall and context injection across connected AI endpoints without explaining that cloud-backed endpoints may receive the recalled local data or requiring informed authorization for that transfer. Calling the memory local and user-owned can obscure this trust-boundary crossing, potentially exposing private preferences, project information, or explicitly stored sensitive information to external providers. The prohibition on storing secrets and relevance filtering reduce exposure but do not replace destination-specific disclosure and authorization; there is no evidence of deliberately malicious exfiltration.

Content

No source excerpt is available for this finding.