Johann Rehberger

• Log-To-Leak: Prompt Injection Attacks on Tool-Using LLM Agents via the Model Context ProtocolVerdict: Useful

  Oct 26, 2026 · A security researcher who has extensively documented prompt-injection data exfiltration in agent clients, the same class of attack.

• V$^3$AST-Bench: Verifiable, Valid, and Versatile Safety Evaluation for Coding AgentsVerdict: Important

  Oct 26, 2026 · Security researcher known for prompt-injection and coding-agent exploit work in the same threat space.

• Combining Long-Term Memory and LLMs Comes at a High Cost: From Prompt Injection to Persistent CompromiseVerdict: Useful

  Oct 20, 2026 · Co-author of SpAIware, the prior memory-injection attack this work builds on and tries to reproduce

1. embracethered.com/blog/about
2. github.com/wunderwuzzi23

Verified Oct 29, 2026