A Substring Is Not a Speech Act: My AI Agent Executed Questions and Quotes

A small language-binding bug let questions, negations, and quoted text trigger an AI agent's actions. The repair replaced substring matching with a finite authored grammar and a whole-response fail-closed check.

Read Original

Related

Dev.to tutorial 49m ago

Stop the action, not the injection

2026's agent attacks are one bug. The fix isn't a smarter model — it's a signed limit at the tool call, outside the agent. Four runnable demos.