Do I need to parse my existing RD?

[streeyt]

Hi,

Assuming I already had a fairly well crafted PRD in MD that I'd spent time brainstorming and producing with Claude.

Is it strictly necessary to tell Task Master to parse that document and recreate a fresh one based on the example-PRD.md that ships with Task Master?

I'm just thinking that I'd be taking a PRD that I'm happy with, having AI process it again only to come up with a flavour of the original work that in my mind was fine in the first place.

Thanks

[eyaltoledano]

Good question. Short answer: nope. You can just use your mega-detailed PRD.

The biggest thing our example PRD does is that it enforces the idea of dependency aware and logically sequenced tasks. You want to make sure the PRD is describing the sequence in which implementation should be done.

This helps avoid situations in which you start building Core Feature A, get to 50% of it, realize it depends on Core Feature B, and then start working on Core Feature B with Cursor getting confused about what to do and getting side-tracked doing Core Feature A stuff.

You want to be doing one task / sub-task at a time to keep the AI on track. And you can only really reliably do that if the dependencies are ironed out, and those will only be useful if they're logically thought through by the LLM. The sequencing helps make it so each task builds on the last and you can go in sequence from task 1 to 2 to 3 to 4 rather than going from task 1 to 8 to 3 to 2 to 6 due to dependencies. You might still run into that, but we try to avoid it.

If your PRD is strong, just make sure to add the section from ours that describes the above and you should be good to go. Just parse yours directly.

Let me know how it works for you!

[streeyt]

Cool, thanks for the great explanation.

[flipbytes-dk]

If I have an already existing project and now need to implement some additional features/fix bugs, can I put them in a PRD and ask task-master to do its magic?