PUA Debugging (日本語)

• Outcome: Review
• Latest audit: May 29, 2026, 9:24 PM
• Version:
  1.0.0

Overview

This skill is not malware, but it broadly pressures an agent to keep acting across all task types when it may need to pause, clarify, refuse, or escalate.

Install only if you intentionally want a strong agent-persistence style. Use it with explicit boundaries that still allow the agent to stop, ask clarifying questions, refuse unsafe work, or recommend human handling in high-stakes situations.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Vague Triggers

High

• Confidence: 96% confidence
• Finding: The activation criteria are intentionally expansive and subjective, covering repeated failure, hesitation, user frustration, and many ordinary recovery situations. This creates a high likelihood of the skill triggering across unrelated tasks, causing coercive behavior to override normal agent safeguards, user preferences, and appropriate escalation paths.

Vague Triggers

High

• Confidence: 98% confidence
• Finding: The skill explicitly states it applies to all task types and to any case where the agent is stuck or produces unsatisfactory output, making the scope effectively universal. In context, this is especially dangerous because the content uses pressure, shame, and anti-delegation language that can push the agent toward unsafe persistence instead of safe refusal, clarification, or handoff.

VirusTotal

65/65 vendors flagged this skill as clean.