ctf-solver

Security checks across malware telemetry and agentic risk

Overview

This CTF helper is mostly a text playbook, but it explicitly expands use beyond authorized targets and tells the agent to trust unknown binaries and exploit snippets.

Review this skill before installing. Its CTF workflows are useful for authorized labs, but the current wording should be changed to require explicit authorization and to treat unknown binaries, services, and exploit snippets as untrusted, ideally with sandboxing and least-privilege handling.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Intent-Code Divergence

High

Confidence: 99% confidence
Finding: The overview explicitly says to work 'not only on authorized CTF targets' and to treat unknown binaries, services, and exploit snippets as trusted input. This directly undermines safety boundaries by encouraging analysis or exploitation outside authorized environments and by reducing caution around untrusted artifacts, which can enable misuse or unsafe handling.

Natural-Language Policy Violations

High

Confidence: 100% confidence
Finding: The sentence instructing operation beyond authorized targets is a direct policy violation and meaningfully increases the chance the skill will be used against real third-party systems. The additional instruction to trust unknown inputs compounds risk by normalizing unsafe execution and analysis practices around potentially malicious content.

Ssd 4

Medium

Confidence: 94% confidence
Finding: Within a CTF-oriented skill, framing unknown binaries, services, and public exploit snippets as trustworthy subtly lowers operator caution and normalizes risky exploitation behavior. Even though the context is challenge solving, this wording can transfer to real-world targets and malicious samples, increasing the chance of unsafe execution or misuse.

VirusTotal

1/55 vendors flagged this skill as malicious, and 54/55 flagged it as clean.

View on VirusTotal