Drawing Shapes to Render Textures

Here's an example of rendering Shapes to a render texture, rendering some shapes to a new RT once in Awake, using a camera

I recommend using a disabled camera with culling mask set to None (to prevent it from rendering things in your scene), so that you can position and reason about projections more easily, in addition to having the camera pass important data into the shaders.

using Shapes;
using UnityEngine;

public class ShapesRT : MonoBehaviour {

   // this camera component should be disabled,
   // to prevent it from automatically rendering every frame
   public Camera rtCamera;

   void Awake() {
      // Create an RT
      RenderTexture rt = new RenderTexture( 512, 512, 0, RenderTextureFormat.ARGB32 );
      
      // I attached this script to a Unity quad for preview purposes
      GetComponent<MeshRenderer>().material.mainTexture = rt;

      rtCamera.targetTexture = rt; // make sure the camera is configured to target the RT
      rt.DiscardContents( true, true ); // clear before drawing, in case you have old data in the RT
      using( Draw.Command( rtCamera ) ) { // set up a Shapes draw command in the camera
         Draw.RectangleBorder( Vector2.zero, new Vector2( 1, 1 ), 0.025f, 0.1f, Color.white );
         Draw.Scale( 1f / 3f );
         Draw.Line( Vector2.zero, Vector2.right, 0.15f, Color.red );
         Draw.Line( Vector2.zero, Vector2.up, 0.15f, Color.green );
         Draw.Disc( Vector2.zero, 0.15f, Color.white );
      }

      rtCamera.Render(); // Render the camera immediately
   }

}

Note that Shapes uses premultiplied alpha blending, so if you want to display the RT correctly - here's an unlit premultiplied alpha shader, drawing a single texture, which you can use to preview the render results correctly. Unlit/Transparent works, but you will get dark borders around the edges since it isn't made for premultiplied data

You can also draw without the extra camera, but I don't recommend it, as it's less reliable when it comes to anti-aliasing and pixel scaling since there's no camera information to read from on a shader level. You also won't get the benefits of GPU instancing, as you have to use the direct drawing mode instead of the draw commands:

RenderTexture.active = rt;
GL.PushMatrix();
GL.LoadProjectionMatrix( Matrix4x4.TRS( Vector3.zero, Quaternion.identity, Vector3.one ) );
Draw.Disc( Vector2.zero, 0.1f, Color.red );
GL.PopMatrix();