An AI agent can help you read the project, write code and draft a note. It works best when it starts from the project's own rules, so every route below begins with the same short brief. You sign in to the assistant yourself; this page never sees your account.
Do not paste passwords, tokens or private data into a chat assistant, and do not let an agent do it for you. You own every number and every conclusion in your work, not the assistant. The rules are in AGENTS.md and, for students, in student guide 5.
The brief says what the project is and what an agent must never do (never commit to develop or master, never merge, never invent a number). Copy it now: you will paste it if an assistant cannot open web pages.
Loading... (if this stays empty, open agent-brief.txt)
A chat assistant reads what you give it, but it cannot see your files or run anything. Use it to understand code, draft text and check what you paste in. Each button opens the assistant with a short first message that points it at the brief. If the assistant says it cannot open the page, paste the brief you copied. If it opens empty (some sites drop the text when you sign in), click the button again once you are signed in, or paste the brief.
A caution notice is normal. Claude, and possibly others, show a warning (for Claude: “Use caution before running this prompt”) before they run a message that came from a link. That is their safety check, not an error. Read the first message: it only asks the assistant to read the brief and then ask you what you want to do. If that is what it says, continue.
If an assistant’s page does not load at all, a browser privacy setting or content blocker is the usual cause. Open the site yourself, sign in, and paste the brief.
Gemini has no link that fills in the first message, so its button only opens Gemini: paste the brief you copied.
These agents run on your computer, read the project's AGENTS.md by themselves, and can edit files, run the tests and use Git. First get the project, once (the setup guide does this for you):
git clone https://github.com/villano-lab/NSDF-Data.git
cd NSDF-DataThen start the agent you use, from inside the NSDF-Data folder. A student first creates a student-yourname branch (setup guide, Step 8) and works only there.
It reads CLAUDE.md, which points to AGENTS.md.
claudeIt reads AGENTS.md.
codexIt reads GEMINI.md, which points to AGENTS.md.
geminiOpen the NSDF-Data folder (File, then Open Folder) and start the agent or chat panel. They read AGENTS.md from the folder. Check the first reply: if the agent does not mention the project's rules, paste the brief you copied.
A codespace is a computer in the cloud with the project already set up (Python 3.12, the pinned packages and the Jupyter kernel). You work in your browser, with the files and a terminal, and you can run an agent there too.
You need a GitHub account. Codespaces uses free monthly hours, so stop the codespace when you finish. Student guide 6 walks through it. This route is still being tested: the environment is checked, but the data download, the unit tests and the quiet-trace check have not yet been run inside a codespace.
Agents that browse the web can start from llms.txt (a short index of the project) or directly from the brief. The complete rules are in AGENTS.md.
Logos belong to their owners and appear only to identify the service a link opens. Most come from Simple Icons (CC0). Where that set has no logo (ChatGPT and Codex, Microsoft Copilot), a plain symbol stands in. Details: img/agents/README.txt.