I use sqlldr on a monthly basis as follows:

1. Create a control file - filename.ctl
2. Create an executable script - load_file (or whatever you want to call it)
3. Make sure the file you want to load does not change name or type, mine is just called start.csv and in comma seperated format. Write some error checking code for this.

##filename.ctl##
load data
infile start.csv
into table contact append
fields terminated by ',' optionally enclosed by '"'
(
URN,
TITLE,
FIRST_NAME,
OTHER_NAME,
LAST_NAME
)

##load_file##
sqlldr scott/tiger control=filename.ctl

##start.csv - first record example##

"0061031","Mr","B","A","Williams"

load_file will look at the control file and import the values from start.csv as required. If you sure the file is being placed in a directory it is worth while checking whether it exists before trying to import it into oracle else abort.

sqlldr does create a filename.log in the directory where you specified unless you specify a path as shown with the example by vgersh99.

Pretty basic but with a cron job should be able to get it working easily.

Hope this helps.