In the end I dodged editing the file on the hard drive by attaching the HD to another VM. Instead I changed the Linux install disk's boot configuration to include a hard-coded data including the path to the anaconda kickstart script, and edit the kickstart script before starting the Linux boot.
Some details:
unpack the CentOS boot CD ISO (mount the ISO file with -o loop option, copy everything to a new directory)
edit the file isolinux/isolinux.cfg to add this new default boot command:
append initrd=initrd.img ramdisk_size=9216 ks=ftp://10.200.100.182/kickstart/ks_alan.cfg ip=10.200.100.33 netmask=255.255.255.0 gateway=10.200.100.1
use mkisofs tool to make a new ISO file
put the new ISO in the ESXi host's datastore
The obvious disadvantage is that the ISO has to be re-made when our build machines' IP addresses change, but that's very rare for us. If it ever happens then maybe we'll bite the bullet about attaching the HD to a new VM.
All the rest of the system configuration is done by the kickstart script: Linux core is installed; our appliance is installed on top; versioning and other changeable data is pulled in by anaconda from the kickstart script definitions.
Thanks again for the nudge in the right direction.