仕事

Keep context in GitHub + Notion, not locked in one lab

Grok Bot is great until your memory only lives inside one vendor.

困っていたこと

Khe Hy (@khemaridh) has been heavy on Grok Bot but does not want context trapped there. He keeps durable notes in GitHub: routines and skills convert files to markdown, add metadata, and file them — mostly agents touching the files. For structured work he uses Notion databases (pipelines he will actually edit) and Notion task lists various bots read. Connectors stay on his side so he can plug the same context into different LLMs.

インストールまたは作成するもの

GitHub as markdown memory via routines/skills. Notion as structured DBs + bot task lists. Bots read and write there; you own the store.

ボット

Context Clerk
Convert agreed file drops to markdown, add metadata, file into GitHub folders. Mirror structured tasks into Notion DBs you named. No silent store deletes.

プラグイン

Notion
必須
ブラウザ
任意

ファイル

A GitHub repo (or folder) for markdown memory + a Notion DB for pipelines/tasks.

ルーティン

Periodic convert-and-file pass into GitHub. Notion stays the human-edited structure layer.

プロンプト

これはヒントであり、貼り付けられるシステムプロンプトではありません。作者はプロンプトを公開していません。初回の手順に従ってください。

初回実行

1Read https://x.com/khemaridh/status/2096003811974529038.
2Pick a GitHub repo for durable markdown. Create a Notion database for pipelines or bot task lists you will actually open.
3Create a Context Clerk bot. Connect Notion (and browser/git access as you use them).
4Hang a routine or skill: convert drops to markdown, tag metadata, file into folders — agents touch files more than you do.
5Keep Grok Bot as the worker, not the only memory. Same stores should stay usable with another model later.