Doing a mandatory #Copilot training and, even if we disregard all the many usual reasons to avoid using LLMs, why would you bother using Copilot cloud agents for GitHub with all these red flags? learn.microsoft.com/en-us/training/modules/github-copilot-code-agent/2-security-risks-limitations-co…
Text: "Compatibility limitations: Doesn't sign commits. If you require signed commits, you must rewrite the commit history before merging. Requires GitHub-hosted Ubuntu x64 runners. Self-hosted runners are not supported. Not available for personal repositories owned by managed user accounts (runners unavailable). Doesn't honor content exclusions; the agent can see and update excluded files. The "Suggestions matching public code" policy isn't enforced by the agent; references may not be provided. Works only with GitHub-hosted repositories. You can't change the AI model used by the agent; it's selected by GitHub."