This project sounds as if you are looking to create an extended watchdog api for a suite of applications. If this is the case, I am quite familiar with this style of application and have created several unique watchdog API's for RTOS systems in the past few years. Probably the most tightly bound was a system with 128k spare SDRAM(original size of ram was 512k with a 300k+ RTOS application based on nucleus). The API I created offered registration, check-in, memory monitoring and recovery of the system with no escapes.
I would also like to recommend a tiered structure where we first settle on a requirement spec, and a milestone based execution structure with you being able to verify each step.
Is there a specific toolchain and/or processor that you are targeting this for, or are you expecting a generic x86 system?