ClawHub
Back to skill

Security audit

anthropomorphic-agent-engine

Security checks across malware telemetry and agentic risk

Overview

This is a local Python psychology-simulation skill with some version-labeling and activation-scope issues, but no hidden, destructive, credential, network, or persistence behavior was found.

Use this only if you want a local Python-based agent psychology simulation. Before relying on it, verify which engine version is actually present, and invoke it explicitly for modeling or implementation tasks because the trigger examples are broad. Commercial or government users should also review the restrictive license terms.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (3)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The file claims it is using a V8.0 engine while it dynamically imports and aliases SPLPureCoreV7_3. This is a supply-chain and trustworthiness problem: operators may rely on advertised safety or behavior guarantees from V8.0 that are not actually present, leading to incorrect risk assumptions, debugging errors, or deployment of an older core with missing controls.

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The runtime banner and final summary present the engine as SPL Pure Core V8.0 even though the code constructs SPLPureCoreV7_3. Because this appears in live output, it can directly mislead users, auditors, or downstream systems into believing newer functionality or safeguards are active when they are not.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The trigger scenarios are broad enough to match ordinary conversations about agents, psychology, games, or behavior design, which can cause the skill to activate outside narrowly intended contexts. Overbroad auto-activation increases the chance of unneeded capability injection, user confusion, and unintended influence over responses, especially in systems where skills alter behavior automatically.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.