Construct 3 isn't a 3d engine. Easiest would be to just fake as much as you can. But here are some thoughts on various ways to do shadows such as "shadow mapping", "shadow volumes" and "ray casting". In construct there are limits but we do have math, distort meshes, canvas and shaders that can be used in creative ways.
Shadow mapping
This is a common way to do shadows where you first render the scene from the point of view of light to a depth texture and then use a shader when drawing from the camera's view to tell what is in shadow.
Construct can only render one view at a time and the canvas has no depth. However you probably could make an effect that takes the depth and draws a color. Then you could change the view from the light, use the effect to draw the depth as colors, snapshot the screen, then use that texture with another effect to draw the shadows.
- you need to have the camera and view matrices. It's not readily accessible but you can calculate them manually.
- you have to write shaders. You are also limited number parameters in C3 shaders so you will have to adapt any online guide about shadow mapping.
- not suitable for Realtime since you have to dedicate a frame to draw the depth.
- on top of that capturing the screen can take a bit.
- Lastly and most importantly It can't work because we can't access multiple textures in shaders in C3. I can't think of a workaround for this.
Shadow volumes
Uses polygon clipping to find what area of each polygons are in shadow. Often done along with BSP trees to be more efficient.
- you need a list of the vertices and faces. C3 does not provide access to those so we'll need to calculate and provide them ourselves.
- BSP trees are needlessly complex to implement in events, but even with them this shadow method doesn't scale as well with higher polygon counts.
- will give perfectly crisp shadow edges, and will be 100% black without further trickery.
- will have to draw the shadows over the mesh with distort meshes. Probably will have to move slightly toward the camera to avoid z fighting.
- This is the closest to being feasible in vanilla C3 solution, but it may not fit in 25 events.
- may only be fast enough for Realtime with low poly counts, but it would have to be made first to find out.
Ray casting
This is an alternate idea to make the 3d objects from canvas' wrapped with distort meshes. The idea is to shoot a bunch of rays from the camera and when the mesh is hit you'd calculate where on the canvas object was hit and color that.
- it would be pretty slow so you'd only be able to shoot a few hundred rays per frame. But over time it would be better shaded.
- It would also not scale well with higher poly counts.
- as with the other approaches it would take a fair amount of 3d math with matrices and vectors.
Overall even without the limits of the 3d in construct those are all fairly involved to do. The second way seems almost doable but likely just as a novel demonstration.