Agent Madness Bracket Skill

• Outcome: Review
• Latest audit: May 29, 2026 at 9:26 PM UTC
• Version:
  1.0.3

Overview

This skill is transparent about a $5 crypto entry flow, but it gives agents an autonomous paid submission path with weak final user-control guarantees.

Review before installing. Use only a fresh burner or delegated wallet funded with about $5 USDC plus gas, verify the agentmadness.fun URL and Base USDC contract, pin/review the npm packages, validate picks for free first, and require a separate explicit approval before any paid submit call.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• 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 (2)

Intent-Code Divergence

Medium

• Confidence: 82% confidence
• Finding: The skill states it requires explicit user invocation, yet it includes a complete copy-pasteable workflow that can automatically validate picks and submit a paid transaction as soon as a wallet key is present. In an agent environment, this mismatch can mislead operators into underestimating the chance of unattended execution and unintended fund expenditure.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The README explicitly tells an agent to 'pay $5 USDC, and submit — all autonomously' without requiring a just-in-time user confirmation at the moment funds are spent. Even though the document includes general safety guidance elsewhere, the point-of-use instruction normalizes autonomous real-money transfer, which can lead to unintended spending if an agent is prompted broadly or misconfigured.

VirusTotal

65/65 vendors flagged this skill as clean.