AI自我意识引擎

• Outcome: Review
• Latest audit: May 29, 2026 at 9:38 PM UTC
• Version:
  1.1.0

Overview

This skill is not malware, but it proposes shared persistent AI memory and cross-agent APIs without enough clarity about access control, consent, retention, or deletion.

Review before installing. Use this only where shared long-term memory is acceptable, and confirm who can call the API, where logs and memory files are stored, whether access is local-only or authenticated, how sensitive data is filtered, and how users can review or delete stored state.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (4)

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill explicitly promotes a multi-agent shared consciousness network with shared event logs and identity-anchor APIs, but it does not disclose privacy implications, consent requirements, or what data may be exposed across agents. This creates a real risk that prompts, answers, predictions, and identity-related state are transmitted or retained in ways users and operators do not expect.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The distillation feature describes automatically writing session-derived insights into long-term memory files without any user-facing warning or retention safeguards. Persistent storage of conversation-derived content can capture sensitive data, personal information, or confidential context beyond the user's expectations.

Ssd 3

Medium

• Confidence: 94% confidence
• Finding: These sections encourage long-term logging of 'self events' such as predictions, reactions, identity shifts, and insights in JSONL, explicitly supporting multiple agents in a shared log. Because the logged data is free-form and conversation-derived, it can easily include sensitive prompts, inferred preferences, internal reasoning traces, or other confidential information without access controls or minimization safeguards.

Ssd 3

Medium

• Confidence: 95% confidence
• Finding: Session-end distillation is presented as an automatic mechanism to preserve lessons from a conversation into long-term memory, but no safeguards are specified for secrecy, consent, or deletion. This increases the chance that sensitive user content is transformed into durable memory artifacts that persist beyond the original session context.

VirusTotal

65/65 vendors flagged this skill as clean.