How an AI agent swarm solved 25 CTF challenges in 24 hours (and what it means for security testing)

Описание и применение

Сwarm AI-агентов успешно решил 25 задач CTF за 24 часа, демонстрируя эффективность параллельной работы по категориям, тайм‑боксинг и устойчивость к сбоям.

Задача

Human teams cannot efficiently solve many CTF challenges within limited time; need scalable, automated approach.

Результат

The swarm achieved 25 flags in ~24 hours, demonstrating parallel category-based automation, effective timeboxing, and resilience to agent failures.

Рабочий процесс

• Deploy 5 autonomous agents on a 7GB VPS
• Assign each agent a category (forensics, web, crypto, etc.)
• Use plain files for shared state (/tmp/ctf_notes_*.md, /tmp/ctf_solves_*.log)
• Timebox each agent to 25 minutes per challenge
• Document every step and log solves
• Merge results via orchestrator
• Handle agent crashes by resuming from notes
• Apply lessons to domain surface scans

Доказательства

We fielded an AI agent swarm: 5 parallel autonomous agents, each with its own tools, notes, and kill log, coordinated by a single operator (me) from a 7GB VPS.

Источник

Result: 25 flags in ~24 hours, across forensics, web, crypto, reverse engineering and misc

Источник

Parallel by category beats parallel by challenge.

Источник

Timeouts kill depth.

Источник