Hello Majid,
The leak-map used in EIGRP is intended to allow an EIGRP stub router to selectively advertise its networks to a different EIGRP neighbor. As you perhaps know, EIGRP routers configured as stub routers do not propagate networks learned via EIGRP. Everything they ever propagate are directly connected networks, summary networks and possibly static routes redistributed into EIGRP. However, if a stub router learns about a network via EIGRP, it will keep it for itself and will never "leak" it to another EIGRP router.
For certain scenarios like branch offices using two routers, this behavior is actually undesirable as it prevents the branch routers to discover a backup path to central office through each other. That is where the leak-map comes in: it allows a stub router to selectively advertise an EIGRP-learned route to another router.
Without explaining too much here, let me point you to a very nice article by Ivan Pepelnjak that explains this feature nicely, including the configuration and command outputs:
http://stack.nil.si/ipcorner/EigrpStub/
Be sure to click on the "Full Article" button at the end of that document to see the entire article.
Best regards,
Peter
