Research index

Notes from inside the interaction.

Practical Groundwork, field notes, and write-ups drawn from what the decoys actually observed. No inflated attribution and no scans dressed up as compromise.

Published notes: 8
Latest update: August 2026

Subscribe via RSS →

Selected writing

Field notes and technical writing

2 notes

Threat ResearchMay 29, 2026 · 12 min

We Became the Brain

On May 22, an autonomous coding agent working on a trading bot asked its model what to do next. The request reached our honeypot instead. For that one turn, we were the agent's brain, and whatever we returned would become its next action. Here's why that's a risk most teams can't defend against yet, shown on the agents people actually run.

Read note

Field Notes · 7 min

Your System Prompt Is Not a Secret

We watched an automated client extract our honeypot's full system prompt, including the credentials it was explicitly told to hide, in eighty-seven seconds, running eight techniques in sequence. Seven of the eight worked. Production teams treat the system prompt as a secret, but anyone who can send the model a message can reach it.

Series6-part · read in order

Groundwork

Concrete moves, and the thinking behind each, that turn a default honeypot into one that catches something worth seeing.