All MicroEvals
I'd like to create a skill I can run on a fairly routine bas...
Create MicroEval
Header image for I'd like to create a skill I can run on a fairly routine bas...

I'd like to create a skill I can run on a fairly routine bas...

Prompt

I'd like to create a skill I can run on a fairly routine basis, which instructs an LLM to do a comprehensive "rebuild" on our internal LLM-facing documentation, for the particular area/scope that I give it (e.g. "this repo", or "the yanos project"). The ultimate problem I'm fighting is getting LLM systems to construct their markdown documentation in a way that is a) fully accurate and drift-free, even after a major change, and b) well-structured in a way that will be easy to consume by future LLMs who need to find correct information in the future. This means doing things like breaking up large documentation elements into many different smaller files, potentially with multiple layers of hierarchy and structure, with each smaller chunk being linked by a higher level indexing/summarizing markdown document above it. I don't want any specific references to any specific implementation details as a part of this skill, I want a completely generic prompt that can be used to trigger both a) an analysis of how much refactoring needs to be done, and how "off" the current documentation is, and b) apply that refactoring, being certain to explore deeply and thoroughly to make sure we're now missing anything. An ideal implementation would offer a low/medium/high level of pervasiveness and thoroughness, but if that's too complicated we could start with just high.

Drag to resize
Drag to resize
Drag to resize