Ctf Ai Ml

When you hit an AI/ML challenge in a CTF and need to extract flags from neural networks, break LLM guardrails, or craft adversarial examples, this gives you the playbook. It covers model weight manipulation (negating fine-tuning deltas, merging LoRA adapters), classic attacks like FGSM and PGD, prompt injection techniques, membership inference, and model extraction via API queries. The quick start commands are genuinely useful for inspecting PyTorch checkpoints, safetensors files, and HuggingFace models. Honestly, the referenced markdown files do the heavy lifting here, but the triage section helpfully tells you when to pivot to crypto or reverse engineering skills instead.

skills.sh installs: 8.3K
GitHub stars: 3.7K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add ljagiello/ctf-skills --skill ctf-ai-ml --agent claude-code

Installs into .claude/skills of the current project.

Files

CTF AI/ML

Quick reference for AI/ML CTF challenges. Each technique has a one-liner here; see supporting files for full details.

Prerequisites

Python packages (all platforms):

pip install torch transformers numpy scipy Pillow safetensors scikit-learn

Linux (apt):

apt install python3-dev

macOS (Homebrew):

brew install python@3

Additional Resources

model-attacks.md - Model weight perturbation negation, model inversion via gradient descent, neural network encoder collision, LoRA adapter weight merging, model extraction via query API, membership inference attack
adversarial-ml.md - Adversarial example generation (FGSM, PGD, C&W), adversarial patch generation, evasion attacks on ML classifiers, data poisoning, backdoor detection in neural networks
llm-attacks.md - Prompt injection (direct/indirect), LLM jailbreaking, token smuggling, context window manipulation, tool use exploitation