Kiro Powers are on-demand tool bundles in Amazon Kiro that extend the assistant’s abilities through defined permissions and workflows. Instead of being a single fixed feature, a Power can package actions such as reading files, invoking utilities, or interacting with external resources, subject to the user’s approval and the tool’s scope.
In cyber security, this matters because every added capability expands the trust boundary. If malicious text in a workspace can influence the agent, a Power may be used to read sensitive local data, modify files, or trigger outbound requests the user did not intend. That makes Kiro Powers relevant to prompt-injection attacks, where attacker-controlled content tries to steer the model into unsafe actions. Defenders reduce risk by limiting permissions, reviewing installed bundles, separating secrets from workspace content, and requiring human approval for sensitive operations.