(edited) The proposed SS is based on following basic preferences:

-1 loop at each end
-2, 4 or 5 loops between sheets (SS)
-3 loops between helixes (HH), and between helix and sheet (HS)
-helix length = 2 * sheet length

The recipe only proposes designs where there is a reasonable amount of sheets for each helix (minimum 2 sheets per helix) and a reasonable core potential. Only exact matching of above preferences are proposed, unless for very common structure (e.g. SSHSS), where a solution with extra loops could be proposed (up to 5 loops around an helix).

The SS is archived in note 19. It can be recovered at any time using Reset my SS in note 19 _GS.

1) On a reset of the puzzle, run the recipe and select to design you like (e.g. SHS, 2 sheets and one crossing helix)

Now it's time to hand fold (did you expect to escape hand fold ?!?)

2) tool "Ideal SS" on each helix

3) You might freeze all helices and sheets if you want to use bands

4) Cut in the loops - or use bands - in order to align the sheets and make H-bonds

5) Align the helices along the sheets

6) Quickfix

ver 1.0.11 corrected the length of sheets in order to respect chirality in 85 + test bands
ver 1.1.0 Adding different paterns for choice of direction and chirality, e.g. HHSSS, SSSHH, HSSSH
Note: in SSS, when the number of residues is pair (for a connecting loop of 3 segments) , we can do a good
cake turning left, then right then left ...
If it is uneven, it does not work (has to turn always left or always right).
That's what I tried to correct in this verion
Moreover, I've seen in top solutions that HSSSH can be as good as SSSHH or HHSSS
As chirality has its importance, I made several similar structures but in opposite directions as well.
The default loops are : 3 segments for connecting sheets, 4 segments for connecting Helices (to sheet or helice) and
one segment at each end. From this basic designes, it's possible to idealize chang turns by transforming loop segments to sheets.
The rules are: SS link: 2-3 => turn left (from residue n to n+1 with sidechains pointing bellow)
4 segs= 50-50
5 segs: turning right (=> replacing the 2 sheet ends with loops doesn't solve chirality problem)
HS link (2-3-4 loop length): the sidechain of the forst sheet seg must point away from helix
(use this rule to decide on which side of the plane you put your helix)
SH link (2 loop length): the helix is where the latest S seg's sidechain points to
SH link (3 loop length): the helix goes opposite of the last sheet sidechain.

added soms SS.
Tip: start with a length of a multiple of 5 (and add 5 segments to the given structure), run the recipe and choose your SS.
Ideal SS all sheets and helices.
Then use Blueprint tool for sheets.
If the loop length between helices is too short (1 or 2 would be too short for core filter), or if the link between sheets and helices (or between sheets) is not nice, add a segment in the loop and try Blueprint again.

Freeze the nice structures and loops. Add bands between sheets. Clashing importance to 0.1, wiggle backbones until most sheet loops are ideal (check the filter ideal loop to see on screen).

Use remix to try to fix unideal loops. If it does not work, add or delete residues in the loop and try again. If it doesn't work, Autostructure and try again.

CI 1, mutate all. Wiggle all.Mutate all. Shake all.

If there remain unideal loops, delete of add segments, remix, autostrucure, try again or run Find Starting Filter recipe if you have no time.

Adapted for binder designs (only set SS for unlocked segments if unlocked are on last segments)

Thanks Bruno :) This is much better than our old starter recipes which didn't address ideal loops at all

Adapted to most common Blueprint shapes (3 loops around helices). Less loops around helices give non-parallel structures, which is less likely to be ok for monomer core filter.

When I try something that immediately fits exactly the filters constraints for monomer core and ideal loops, I add an "ok bp" as a comment within the recipe (this can be usefull for future updates, e.g. indicating "ok bp" in the dialog when appropriate).

It only works curently for number of residues as multiple of 5. For lengths in the between, user could edit the recipe and copy the proxy SS into text editor then adapt and use recipe SS Edit or AA Copy Paste Compare in order to input the SS into the puzzle.

love it. I tried all of them against covid 15 and HSSSSH would shake and wiggle right into a desired shape (all of the sheets eventually bonded nicely w/o any pulling on my part.) And it scored rather high w/ no intervention.

It makes me wonder what was so special about that pattern that would lend it to naturally going into a shape I would imagine.

And it would also be nice to throw many others in to see how they'd perform.

HSSHSSH

I also try to play with shapes like a Greek Key Motif https://en.wikipedia.org/wiki/Structural_motif and to see what it scores like

I think having a loop at the end of your beginning or end sheet is not a good idea. It is very hard to get ideal loops because Foldit thinks the loop is a sheet and wants it to bond to the sheet next to it.

Automatic ideal SS at end of process.

Now what you still need to do:

-use Blue Print in order to idealize the loops
-verify the core monomer filter: if it's too low, freeze backbones, unfreeze some strategic loops, add bands in order to compact the protein, clashing importance 0.9, wiggle backbones untill the core monomer filter is ok. Freeze all backbones, clashing importance 1. Put the protein at right place (near to symmetrics or binding partner), mutate, shake etc.