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

Description and use

An AI agent swarm solved 25 CTF challenges in 24 hours, showcasing category‑based parallelism, timeboxing, and resilience to agent failures.

Problem

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

Outcome

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

Workflow

• 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

Evidence

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.

Source

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

Source

Parallel by category beats parallel by challenge.

Source

Timeouts kill depth.

Source