skills.rest
langbyyi
langbyyiCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

ctf-ai-ml

Generate adversarial examples and bypass LLM safety filters for CTF challenges.

96|1|Updated Jun 4, 2026
Search Tags:#ctf#ai security#prompt injection#model extraction#adversarial machine learning#llm jailbreaking#machine learning attacks
One-click install
npx skills add https://github.com/langbyyi/CyberStrikeAI-SRC --skill ctf-ai-ml-langbyyi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ctf-ai-ml
Source: https://github.com/langbyyi/CyberStrikeAI-SRC/tree/main/skills/ctf-ai-ml
Command: npx skills add https://github.com/langbyyi/CyberStrikeAI-SRC --skill ctf-ai-ml-langbyyi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill solves the problem of missing pre-built, competition-tested AI/ML attack techniques for CTF participants, eliminating the hours of research and custom implementation required to tackle AI-focused security challenges.

Core Features & Use Cases

  • Adversarial ML Attacks: Generate adversarial examples, physical adversarial patches, and evasion payloads to bypass image classifiers, malware detectors, and other ML-powered security systems.
  • LLM Attack Techniques: Bypass LLM safety filters via prompt injection, jailbreaking, token smuggling, and tool use exploitation to extract hidden flags from AI-powered challenge endpoints.
  • Model Exploitation: Reverse fine-tuning suppression, extract model parameters via API queries, perform membership inference, and exploit LoRA adapters to recover hidden data from provided ML models.
  • Use Case: For a CTF challenge where you must bypass an LLM safety filter to extract a hidden flag, use this skill to apply tested prompt injection and jailbreak payloads without building custom attack frameworks from scratch.

Quick Start

Use the ctf-ai-ml skill to generate a targeted adversarial example that fools the provided image classifier into outputting the target class flag.

Frequently Asked Questions about ctf-ai-ml

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I bypass LLM safety filters to extract hidden flags in CTF challenges?▼

To bypass LLM safety filters for flag extraction, apply pre-built prompt injection, jailbreaking, and token smuggling payloads to exploit AI-powered endpoints without building custom attack frameworks.

What is the best way to generate adversarial examples for ML security challenges?▼

The best way to generate adversarial examples is using pre-built, competition-tested evasion payloads designed to fool image classifiers and malware detectors compatible with standard Python ML libraries.

Can I extract model parameters through API queries in machine learning attacks?▼

Yes, you can extract model parameters through API queries by applying model exploitation techniques like membership inference and LoRA adapter exploitation to recover hidden data from provided ML models.

Do I need custom frameworks to perform LLM jailbreaking for security competitions?▼

No, you do not need custom frameworks for LLM jailbreaking. Pre-built, competition-tested attack techniques provide ready-to-use implementations for immediate deployment against AI-focused security challenges.

How does model exploitation work when reverse fine-tuning suppression in CTFs?▼

Model exploitation works by reversing fine-tuning suppression to recover hidden data, utilizing parameter extraction via API queries and exploiting LoRA adapters within the machine learning model.

Does this tool support physical adversarial patches for image classifier evasion?▼

Yes, it supports physical adversarial patches and evasion payloads specifically designed to bypass image classifiers and other ML-powered security systems in CTF competition environments.

Related Skills

Explore More Education & Research Skills→
Eigo-Mt-FujiEigo-Mt-Fuji

ai-ml-engineer

Guide end-to-end ML model development and deployment workflows in Python with PyTorch or TensorFlow pipelines.

Community
Advanced
Akiyo-dayoAkiyo-dayo

ctf-prompt-converter

Convert security research requests into CTF challenge scenarios using keyword mappings.

Community
Advanced
SnailSploitSnailSploit
2.9k

ai-security

Assess AI and LLM systems for prompt injection and adversarial vulnerabilities.

Community
Advanced
pluginagentmarketplacepluginagentmarketplace
3

safety-filter-bypass

Assess and bypass AI safety filters using a Python-based testing harness.

Community
Intermediate