Indirect Prompt Injection in Issues, PRs and Docs | Guide

How content an AI agent reads, such as issues, pull requests, emails and web pages, can steer its actions, and how to design safe trust boundaries.