Save chat summaries in a repo note file
Use ChatGPT with GitHub access to turn long Q&A sessions into stored markdown notes inside the right repository.
- What they did
- They connected ChatGPT to the GitHub MCP server and gave it access to several of their repositories. Each repo included an AGENTS.md file, and a custom instruction told ChatGPT to check the relevant one when a session began or when asked. After a long discussion, they had it write a summary into the chosen repo, using the file locations and note types listed in AGENTS.md.
- How it went
- The setup took effort at the start, but it made the assistant more dependable than trusting its own memory. The notes could capture discussions, decisions, and research ideas in a place the user could come back to later.
- Tips
- Set up repo access, AGENTS.md guidance, and a session-start instruction before you rely on this workflow. Make the agent show the target file and contents first, then wait for approval before any commit or other irreversible action.
Try it yourself with ChatGPT
You are connected to my GitHub repo [owner/repo]. First, read the repo’s AGENTS.md and follow its instructions for where discussion notes should go. After we finish talking about [topic], draft a concise markdown summary covering the question, the options considered, and the decision made, and place it in [notes folder path] as [file name]. Before you create, commit, or send anything, show me the exact path and the full draft contents and wait for my approval. If I approve, save it to the repo and then stop.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Use AI to map a thesis literature review
They used ChatGPT to summarize papers and sketch the review, then switched to a source-first tool to cut citation risk.
Turn a conference website into a searchable schedule dashboard
A Manus agent scraped a conference site, normalized talk times to UTC, and assembled a dashboard with search, recommendations, and calendar export.
Search large public archives with plain language
Claude Code can translate everyday questions into SQL or vector queries for Scry's read-only data index.
Turn Slack research into a Notion page draft
Lindy reviewed past guides and a model in Slack, then prepared a Notion idea and waited for sign-off before making it.
Use a visual browser to compare online stores
A commenter used ChatGPT Agent as a visual web browser to check many stores for one item and see which ones would ship to their country.
Use Claude Code to coordinate research and drafting
Researchers used Claude Code to manage lab-style work and produce papers or theses in LaTeX or Markdown.