This collection provides techniques for navigating the intersection of artificial intelligence and security competitions. It enables agents to perform model extraction, craft adversarial examples, and execute prompt injection or jailbreaking attacks.
The skill set is designed for CTF players automating the analysis of neural networks, LoRA adapters, and LLM vulnerabilities to uncover hidden flags or manipulate model behavior.
Follow the documented setup, then try a first task.
Installs: Complete ljagiello/ctf-skills repository bundle; skillPaths is a bounded selection of positively supported paths, not the complete bundle inventory. · Any tool supporting the Agent Skills specification, including Claude Code
In a terminal, add the complete ljagiello/ctf-skills repository bundle.
npx skills add ljagiello/ctf-skills
Uses solve-challenge
/solve-challenge <challenge description or URL>
Uses ctf-writeup
Use the ctf-writeup skill to draft a concise submission-style writeup for this solved challenge. Include the supplied metadata, one reproducible solution path, the smallest complete solve script, and the final flag. Do not publish it.
Uses ctf-malware
Use the ctf-malware skill to perform static triage of the suspicious file at <sample-path>. Identify its file type, notable strings, and possible network indicators. Do not execute the sample.
Claude
Antigravity