← sources

Someone here open-sourced skills so project decisions and handoffs can follow you between Claude, Codex, Gemini and Grok. How do you handle switching assistants mid-project?

source: reddit (reddit:ClaudeCode)published 2026-11-19 11:46ingested 2026-11-20 02:39

retained text

On Nov 16 a post in this sub described two open-source Agent Skills: Repo Memory, which saves a project's decisions and progress, and Repo Relay, which passes tasks, questions and handoffs between agents. The author lists Claude Code, ChatGPT/Codex, Gemini CLI, Grok Build, Cursor and supported Copilot agents, and says they have not tested every platform, so I would read that list as intent rather than a guarantee.

The reason they gave is the interesting part. When you switch assistants mid-project, you end up re-explaining what was decided and what is still unfinished.

I can't tell how common that is. Plenty of people probably pick one assistant per project and never hit it. So I am curious how it looks from your side:

* Do you switch between assistants on the same project, or stay with one?
* When you do switch, what carries the context over: a file in the repo, a handoff note, something else?
* Has any setup like this lasted more than a week without turning into more upkeep than re-explaining?

Link to the original post is in the first comment.

document fetched 2026-11-20 00:41 · canonical https://reddit.com/r/ClaudeCode/comments/1y07e70/someone_here_opensourced_skills_so_project