question: regarding usage of PRD file

[Atheer2104]

I have a question regarding the PRD file what are supposed to do with it after it has been generated and task has been generated using it, is no longer needed ?

Furthermore if we are adding new tasks, should the PRD file be updated or can we just add them directly ? currently I have added them directly and one thing that I'm noticing is that these new tasks do not get dependencies, maybe this can be added by analysing the new tasks (not tried) or ?

[eyaltoledano]

Currently, once you've parse the PRD, there's no use for it. You can then parse additional PRDs to append to an existing list of tasks (with --append flag on parse-prd) just make sure the subsequent PRDs contain only new information.

[BradKML]

Are we obligated to add details to tasks after parsing the PRD? Cus sometimes when the PRD receives new information, some of the original details gets deleted since file modification is often not done with diff incrementally

[Crunchyman-ralph]

@BradKML don't understand your question.

[Crunchyman-ralph]

@prime-optimal this is an MCP client limitation of the AI editors where they don't allow us to report progress. But per the mcp protocol, there is a possibility of showing progress and letting the user know what is going on, but right now the MCP clients haven't implemented it.

But I'm sure it'll be available soon, mcp is so new and we're pretty early

[BradKML]

@Crunchyman-ralph as a weird observation, sometimes when PRD gets rewritten, it will displace old details in favor of whatever is mentioned, so line-by-line it stays somewhat the same. Kind of a weird quirk of LLMs in general to not be able to do longforms.

[eyaltoledano]

Just wanted to say that it took me a while to figure out that if I used --append, I didn't need to also pass --input anymore.

Also, it would be great if there was a verbose mode or some sort of way to see exactly what's going on in these situations. I dunno if it's just claude 3.7 being verbose or if the models are slow right now, but it's frustrating to wait forever only to have an MCP error or OpenRouter error at the end

@joedanz has been cooking something to show this in the CLI but as ralph mentioned, we actually already report progress during the parse PRD process - the issue is that most MCP clients aren't consuming these updates so it looks like it's just hanging there doing its thing.

The CLI of course is a different story, and Joe's improvements basically log what's happening during parse-prd as it's happening