Microsoft has integrated agentic Artificial Intelligence features into the Windows 11 taskbar that let users summon and run agents from the Ask Copilot search field. Users trigger an agent by typing a prompt, selecting an agent from a list, and specifying the desired outcome. Once a task starts, the agent appears on the taskbar like any other app. Hovering over that icon shows a compact status window with progress and any requests for user input. Visual badges indicate when an agent needs input or when a task is finished.
The company says the feature relies on Agent connectors and a Model Context Protocol (MCP) to give agents a consistent way to access applications, files, and services on the PC. Agents run inside an isolated Agent workspace that enforces policies, logs activity for auditing, and is designed to prevent disruption to the active desktop session. Outputs from all agents are routed into the Microsoft 365 Copilot app so users can view results in one place, even when an agent comes from a third party.
Security concerns have emerged alongside the new functionality. The article highlights cross-prompt injection attacks as a significant risk, where attackers embed malicious directives in ordinary documents or interface elements and thereby hijack an agent’s original instructions, causing it to perform unintended actions. Microsoft acknowledges these vulnerabilities in the rollout notes cited and the report urges caution. The article concludes by noting the need for Microsoft to address these security gaps before agentic features receive a wider push.
