By default EIGRP uses a classful network, meaning that it assumes that an address in the class A range with have an 8-bit subnet mask. There is no automatic process of checking other than the assumption that if you do not specify the subnet mask in wildcard mask format ( host mask ) then you want to use the default subnet mask for the IP range you've specified.
As @Todd Wilcox pointed out below, each address range has a class, originally specified in RFC791 ( https://www.ietf.org/rfc/rfc791.txt ), excluding special use addresses ( https://www.rfc-editor.org/rfc/rfc3330 ).
Class Sig Bits Network Mask Range
A 0 8 0.0.0.0 - 127.255.255.255
B 10 16 18.104.22.168 - 22.214.171.124
C 110 24 192.0.0.0 - 126.96.36.199
D 1110 undefined 188.8.131.52 - 184.108.40.206
E 1111 undefined 240.0.0.0 - 255.255.255.255
So converting the first octet ( first group of decimal digits in an IP address: 192.168.1.0 ) into binary gives us 11000000, matching the MSB ( most significant bits ) of a C class address, implying a default subnet mask of /24 bits or 255.255.255.0.