To run the os level jobs you have to set your prefered credential for the node to be the same as the oracle user on the node. Then for the database, you have to set the prefered credentials as system/passwd as SYSDBA. Then make sure that the oracle intelligent agent is running on your node, if not start that. The way you start it on the unix
is

lsnrctl dbsnmp_start;

check the status by dbsnmp_status.

Once you do this, try creating a simple os job of ls and check to confirm that you are getting the out put. Then on the next step, create a simple sql script select * from v$instance; in the script text of the job wizard, and try executing that and see, to make sure that you are getting the output. You could see the output by right clicking the finished jobs and choose the view.

One other question, did you do a patch installation on the database where the managent server is running, on that case, go ahead and upgrade your repository.

If you further have any problem let me know.

Sam