GitHub’s Agentic autofix now uses Copilot Memory for customers who have enabled it. When resolving a security alert, Agentic autofix reviews existing memories and draws on context that may help with the fix. Source
How memory is used to resolve security alerts
Agentic autofix creates fixes for security alerts. With this change, it can use information previously saved in Copilot Memory while working on a fix. A memory is previously saved information; Agentic autofix consults the parts that provide useful context for resolving the alert. This applies to customers who have enabled Copilot Memory. Source
When Agentic autofix creates a fix, it also saves the fix pattern as a memory. A fix pattern is information saved from the fix that can be used when resolving other security alerts later. The change covers both consulting existing memories and saving new fix patterns. Source
Where saved fix patterns can be used
Saved memories can also pass repository-specific secure development patterns to other GitHub Copilot features. The announcement names Copilot code review and the Copilot cloud agent as examples. Repository-specific patterns are approaches to security fixes developed in that repository. Source
Agentic autofix and Copilot Memory are both in public preview. Source