Contribute
Suggest a new page
Write a page, choose where it belongs and label it. It's sent for review, and once approved it appears on the map with the rest.
Prefer to work with an AI coding agent? There's an easier way to contribute ↓
Want an easier way to contribute?
Let an AI coding agent do the heavy lifting
The whole knowledge base is plain files in the manna-discussions repository: one markdown file per page in its docs/ folder. Open that repository in an AI coding tool and every Manna Discussions page becomes context for your agent, so it can answer questions from what's already here and build on it.
The repository also tells agents how to contribute. Its AGENTS.md guide and skills let your agent research and write a new question or a lens from your tradition, label it, connect it to related pages, and check it, ready for you to review and send as a pull request.
- Copy the prompt belowIt sets up a chatbot to be your step-by-step guide.
- Paste it into the chatbot of your choiceClaude, ChatGPT, Gemini, Copilot: any of them works.
- Answer its questionsIt asks about your computer, the coding tool you want and your experience, then guides you from there.
Setup prompt
I want to contribute to Manna Discussions, an open knowledge base about big questions (worldviews, religion and philosophy). The whole knowledge base is a folder of markdown files in this GitHub repository: https://github.com/kalebdixon13/manna-discussions Contributors work with an AI coding agent (for example Claude Code, OpenAI Codex, Cursor, GitHub Copilot or Gemini CLI) running inside their own copy of that repository. Every page in its docs/ folder becomes context for the agent, and the repository's AGENTS.md file and skills (in .agents/skills/) tell the agent how to research, write, label and connect new pages. Please be my step-by-step guide to getting set up and making my first contribution. Before you give me any instructions, ask me these questions one at a time, and wait for my answer to each: 1. What computer and operating system do I use (Windows, macOS, Linux, ChromeOS)? If I'm not sure, tell me how to check. 2. Which AI coding tool would I like to use? If I don't know, briefly compare a few (what each costs, whether it runs in a terminal, an editor or the browser) and recommend one for my experience level. 3. How much technical experience do I have: (a) I've never used a terminal, (b) I've used a terminal a little, or (c) I'm comfortable with Git and the command line? 4. Do I already have a GitHub account, and is Git already installed? Then use my answers to walk me through these steps. Give me ONE step at a time, explain each command in plain language suited to my experience level, and wait for me to confirm it worked (or paste an error) before moving on: 1. Install Git for my operating system and check it works (git --version). 2. Create a GitHub account if I need one, and set my name and email in Git. 3. Fork the repository on GitHub, then clone my fork to a folder that's easy to find. 4. Install the AI coding tool I chose, sign in, and open it in the cloned folder. 5. Follow the "Agent setup" section of the repository's README so the skills are available to my agent. 6. Start working. Ask my agent to read AGENTS.md first, then help me try one of these: - ask a question answered from the existing pages, e.g. "What do different traditions say about the chief end of man? Use the pages in docs/." - answer a new question or add a lens from my own tradition with the worldviews-answer skill - label and connect pages with the brain-map skill 7. Before submitting, make sure the agent ran "npm run map:check" in the front-end folder with 0 errors, and that each new page's frontmatter honestly lists me in "authors" and says how it was written: "authorship" is human, ai-assisted or ai, and "review" says how much of any AI-written text I personally checked (none, partial or full). AI-written text must never be marked as reviewed unless I really checked it. 8. Commit my changes on a new branch, push them to my fork, and open a pull request on GitHub so a maintainer can review it. Ground rules: keep each message short, check in with me after every step, help me read any error message, never ask for my passwords or tokens, and ask me whenever a choice is mine to make.