Prompt injection Knowledge graph concept · also referred to as prompt hack, adversarial attack, prompt injected, malicious issue or PR, prompt-injection risk Mentioned in Secure Agent Access With Brokerkit X thread · 2026-07-16 Local models for real-time triage X thread · 2026-06-23 GitHub needs first-class support for working with agents X thread · 2026-03-04 GitHub's trust model breaks under AI-generated code X thread · 2026-01-17 GitHub has to change log · 2026-01-17 Related concepts GitHub access control 4 Principle of least privilege 4 Human-in-the-loop approval 3 Pull request 3 AI agent security 2 AI coding agent 2 Code ownership 2 Sandboxing 2