I Tried to Prompt-Inject My Own Agent Engine. It Didn't Work. Here's Why.
This is article 5 in a series about building PlannerCritic , an open-source engine where one LLM writes a plan and a second LLM reviews it. Article 1 covers the 157-goal field test. Article 2 is about the critic severity bug. Article 3 is about the planner capability gap. Article 4 is about field test engineering. This one is about what happened when I tried to break my own engine — and what I…
In this article, the author explores why their open-source engine, PlannerCritic, successfully blocked various adversarial prompt-injection attempts. The engine consists of three layers that make direct injection structurally difficult: deterministic gates, the critic evaluating structure rather than intent, and fail-closed abort paths. These mechanisms ensure that even hand-crafted adversarial goals and real-world vulnerabilities, such as those from the SWE-bench security oracle, are blocked.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
