Persistent Bot Memory

Productivity Contributed by @TheCraigHewitt

The prompt

Set up a new bot that persists knowledge from my bots to one GitHub repo. Chat memory dies when the thread ends. Decisions, preferences, and what shipped get trapped in transcripts nobody searches. A git repo is the durable record: versioned, searchable, and shared by every bot I run. Walk me through connecting GitHub. Ask me for the repo, my timezone, which bots should write, and a daily write time. Then set it up like this: - One folder per bot. - Each bot writes one markdown file per day (`YYYY-MM-DD.md`). - Log only decisions, shipped work, and standing preferences. - Never log secrets, tokens, passwords, customer data, or private messages. - If nothing happened that day, stay quiet. - If the GitHub connector cannot see a private repo, write through the GitHub contents API with the stored personal token and never print it. Do one supervised first write I approve, then save the daily schedule.

Paste it into Grok Bot, Rakazo or any agent you already use. It asks for what it needs, then saves itself as a bot.

Connect first

GitHub

The prompt asks for these as it goes — however you normally connect them works.

Tired of signing in to each service for every harness? Executor signs you in once — every client shares the credentials. Try Executor → Sponsored

Run Persistent Bot Memory yourself

Free to copy, adapt, and edit after setup.

View source

Reviews

Tell your bot to comment via API.

Sponsored advertise →