A letter in cipher from Charles Emmanuel, Duke of Savoy (Wikipedia), to his ten-years-younger nephew, Duke of Geneva and Nemours (Wikipedia), dated Turin, 28 December 1606 is preserved in BnF fr.3789 (Gallica), f.3. (The latter's elder brother whom he succeeded was also Charles Emmanuel but had died in 1595.)
The interlined deciphering allows reconstruction of the following cipher.
Figures with an overbar appear to be assigned to small words and syllables in roughly alphabetical order: 13(avoir), 14(au), 21(bon), 31(ce), 34(ce?), 38(de), 39(du), 51(est), 52(et), 53(et?), 55(en), 63(faire), 68(grand), 83(jour?), 86(je), 87(il), 93(la), 94(le), 95(luy).
Bare figures appear to form a sequel, thus avoiding use of three-digit figures: 68(ment), 69(mon), 89(plus), 91(par), 95(pour), 96(pre), 97(pro).
Figures with an umlaut (two dots above) form a further sequel: 3(que), 8(quil), 15(re), 22(sur), 24(sa), 25(se), 26(si), 29(tant), 31(tion), 32(tres), 35(un), 40(vous).
The handwriting of the interlined deciphering made the reconstruction not so straightforward as it should be. "38 with an overbar" for "de", "93" with an overbar for "la", "94 with an overbar" for "le", "3 with two dots above" for "que", and "m"-like symbols for "e" and "s" were among the first that could be identified. By the time the word "vraysemblable" was read, I was sure I was on the right track. More figures may be identified by studying the handwriting of the interlined deciphering.
A cipher used in letters including one from Abate Scaglia to Charles Emmanuel I on 22 June 1630 is printed in the appendix of Toby Osborne, Dynasty and Diplomacy in the Court of Savoy (2007). Unlike the above cipher, it abandoned use of arbitrary symbols and represented letters by figures (a=30, 32, 96, b=33, c=36, 37, 38, ...). Figure "11" with a caret (^) above it represents "gli." Figures with a plus (+) above represent syllables and small words (01=ve, 02=vi, 53=sempre, 99=va). Figures with an overbar represent names.