Loading video player...
GitHub just launched Agentic Workflows in technical preview — AI agents that run automatically inside GitHub Actions to triage issues, fix CI failures, and open PRs. The team that built Copilot built this. Security researchers have already demonstrated a working prompt-injection exploit against the same pattern. I break down what it actually ships, why the security model has a gap, why sandboxing alone doesn't close it, and what you should do right now, depending on whether you're running public or private repos. #GitHubAgenticWorkflows #ContinuousAI #CICD #AIEngineering #PromptInjection #GitHubActions #DevSecOps #AgenticAI #MLOps #SoftwareEngineering #EngineerInTheLoop #AIAgents #GitHubCopilot #ClaudeCode #SupplyChainSecurity