Super Proactive V2

• Outcome: Review
• Latest audit: May 28, 2026 at 11:24 AM UTC
• Version:
  1.0.1

Overview

This skill is not malicious, but it asks the agent to persist memory and run autonomous recurring work with too little scoping or user control.

Install only if you are comfortable with the agent creating persistent memory files in the workspace and acting on recurring task prompts. Keep secrets and private data out of SESSION-STATE.md, MEMORY.md, and memory logs; review those files regularly; add explicit retention/redaction rules; and disable or gate autonomous background tasks unless you want the agent to act without a fresh prompt.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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

Findings (3)

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill explicitly instructs the agent to run autonomous background tasks and persist state to workspace files without any consent, scoping, or safeguards. In practice, this can cause unintended modification of user data, surprise actions, and accumulation of sensitive context in files the user may not expect the agent to create or update.

Ssd 3

Medium

• Confidence: 97% confidence
• Finding: The WAL guidance tells the agent to write critical details, decisions, corrections, and important context to SESSION-STATE.md before responding, which encourages broad retention of conversational and project data. This creates a natural-language data exposure risk because sensitive user inputs, secrets, or internal reasoning-adjacent details may be stored and later disclosed or misused.

Ssd 3

Medium

• Confidence: 93% confidence
• Finding: The instruction to always search memory before answering, combined with the three-tier memory model and long-term retention language, promotes broad storage and resurfacing of prior user information. This increases the chance that stale, sensitive, or context-inappropriate data will be recalled in later responses, leading to privacy leaks or cross-task disclosure.

VirusTotal

No VirusTotal findings