I have been using Claude code to create some custom Frappe v16 based apps. I have already installed the Skills repo to assist.
I have been finding that occasionally Claude stumbles and has to analyse (and sometimes reanalyse again later because it forgot) to overcome what it sees as an undocumented issue or undocumented way of doing things.
What would be the best way to gather this feedback from an AI agent and provide it to the Frappe team to improve the Skills repo? Raise and issue on Github and put it there or are there other preferred ways to submit this?
The correct approach (after you figured it out)
That way the skill instructions can be updated so the AI handles it correctly next time without reanalysis.