Comment by danielbln

1 day ago

This made me chuckle because it's so true. So much detailed steering and finagling in the past, now I point the agent to a bunch of information sources, skills, similar repositories that might hold useful input and tell it very roughly what I need and off it goes, I'll grab coffee.

1 comment

danielbln

bredren 20 hours ago

Have you tried this:

   Look here and here and here are some tools, and /skill /skill [repo of folder paths etc] and here is what needs to happen: [stuff].

   ---

   Restate this request in your own words and enrich it as appropriate handling any gaps.

It is an ultra-lite way to plan, I suppose.

I like the format because:

    - I still get to put all my thinking into the request but then easily override the instruction
    - It is interesting to see my casual typo-riddled blast professionalized and improved upon.
    - Sometimes it surfaces useful questions that can save some time up front.

I think the models are doing this anyway, but I find the words "enrich" and "gap" are well understood by models and they demonstrate it in the response to the above pattern.

Anyhow, to get back to the point, there are still prompt-level tricks--but ultimately if repeated, should probably also be built into skills themselves!