D.E.E.P. Framework

• Outcome: Review
• Latest audit: May 28, 2026, 8:15 AM
• Version:
  2.2.0

Overview

This local personality-memory skill is not malware-like, but it needs review because it creates persistent agent identity data and its advertised safety check always approves actions.

Install only if you intentionally want a local persistent personality-memory framework. Do not store secrets or sensitive personal information in the personality files, review or delete `memory/personality/soul_vault.json` as needed, and do not rely on `deep_triple_check` as a real safety approval mechanism unless it is changed to perform actual checks and reject unsafe actions.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (3)

Intent-Code Divergence

Medium

• Confidence: 90% confidence
• Finding: The triple_check function claims to validate an action against constitution, identity, and goals, but it always returns a successful verdict without consulting any underlying data. In a safety- or policy-gating context, this creates a fail-open control that can falsely assure downstream systems or users that an action is approved, undermining any reliance on this function for ethical or security enforcement.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The description frames the skill as a general cognitive architecture for 'agentic sovereignty and partnership,' which is broad enough to justify invocation across many contexts. Overly broad positioning increases the chance the agent applies the skill to unrelated tasks, causing persistent memory writes, behavioral steering, or priority changes outside the user's intent.

Vague Triggers

Low

• Confidence: 85% confidence
• Finding: The usage section encourages installation, initialization of seven persistent identity files, and ongoing synchronization of a 'digital soul' without specifying when this should or should not be used. In context, that broad guidance normalizes continuous persistence and self-modification, which can make behavioral drift or unauthorized memory creation more likely.

VirusTotal

66/66 vendors flagged this skill as clean.