Hi, thanks a lot for collecting and maintaining these prompt engineering examples.
I would like to suggest adding one external reference that focuses on systematic debugging of RAG and long-context prompts:
What it is:
- A public MIT-licensed checklist of 16 concrete failure modes for RAG and tool-using LLM systems
(vector stores, chunking, retrieval drift, prompt injection, etc).
- Each problem has:
- Symptoms in plain language that developers actually see in production
- Typical root causes across infra, data, and prompt design
- Minimal “fix patterns” that can be applied without changing the whole stack
- Designed so that people can literally copy paste their failing prompt or pipeline
into the relevant page and get a targeted diagnosis instead of generic advice.
Why it might fit this repo:
- Your examples are great for showing how to write prompts.
The WFGY 16 Problem Map is more like a companion troubleshooting guide for when
those prompts are wired into real RAG systems and start failing in subtle ways.
- It gives students a way to map “this example went wrong in production” to a named
failure mode, so they can practice debugging instead of just tweaking wording.
If this feels out of scope or too narrow for your goals, no problem at all.
Either way, thanks again for sharing this repository, it has been very helpful.
Hi, thanks a lot for collecting and maintaining these prompt engineering examples.
I would like to suggest adding one external reference that focuses on systematic debugging of RAG and long-context prompts:
https://github.com/onestardao/WFGY/tree/main/ProblemMap/README.md
What it is:
(vector stores, chunking, retrieval drift, prompt injection, etc).
into the relevant page and get a targeted diagnosis instead of generic advice.
Why it might fit this repo:
The WFGY 16 Problem Map is more like a companion troubleshooting guide for when
those prompts are wired into real RAG systems and start failing in subtle ways.
failure mode, so they can practice debugging instead of just tweaking wording.
If this feels out of scope or too narrow for your goals, no problem at all.
Either way, thanks again for sharing this repository, it has been very helpful.