Visual guides in Virtual Reality sketching systems help to align strokes and raise accuracy. Guides that appear dynamically where one wants to draw have the potential to both lower the reliance on manual guide activation and improve the sketching experience. We propose novel guide techniques for sketching that exploit eye-tracking to modulate where a grid fragment appears. EyeGuide and EyeConGuide cause this grid fragment to appear spatially close to the user's intended sketches, with one relying on the user's eyes and the other on both the eyes and the hand location. In two user studies, we evaluated the techniques in basic and complex sketching tasks. The results show that gaze-based guides increased sketching accuracy compared to no guides, improved the system usability over manual activation, and were preferred by most users. Our research contributes highly usable techniques to assist accurate sketching and insights into multimodal gaze-contingent sketching.