Morgana Anti Infinite Loop V2

• Outcome: Pass
• Latest audit: Jun 9, 2026 at 2:14 AM UTC
• Version:
  2.0.1

Overview

This is a coherent local anti-loop helper with no network or credential behavior, but it does keep a local loop-history file users should know about.

Install only if you are comfortable with local loop history being saved under ~/.anti_loop/loops.json. If agent actions may contain sensitive prompts, file paths, customer data, or secrets, pass a controlled storage_path, restrict file permissions, or periodically delete the history file. There is no artifact evidence of network exfiltration, credential access, destructive behavior, or hidden automatic execution.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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)

Description-Behavior Mismatch

Medium

• Confidence: 85% confidence
• Finding: The module persistently stores loop fingerprints and action samples in a file under the user's home directory across sessions, despite being presented as a lightweight standalone guard. In an agent setting, actions may contain prompts, task data, file paths, or other sensitive workflow content, so this creates unintended local retention and privacy exposure beyond the immediate anti-loop purpose.

Context-Inappropriate Capability

Medium

• Confidence: 87% confidence
• Finding: The LoopDNA feature performs cross-session behavioral tracking by fingerprinting prior actions and retaining them locally. Even if intended for loop prevention, this broadens the trust boundary and can accumulate behavioral history that may reveal user activity patterns or sensitive prompt content over time.

Missing User Warnings

Low

• Confidence: 83% confidence
• Finding: Writing loop history to a predictable file in the user's home directory without an explicit warning at the write site or an opt-in control can surprise users and silently persist potentially sensitive action traces. In shared or managed environments, this may expose metadata or content to other local processes, backups, or forensic collection.

VirusTotal

65/65 vendors flagged this skill as clean.