research-field-onboarding

Language: English | 中文

Ask an assistant about a field you don't know and it answers at the level of someone who does. This one checks the field's own structure, asks which of its prerequisites you already have, teaches one step per turn, and never names a paper it hasn't checked.

🆕 New in v1.9 — fewer questions up front, and one part of the field per turn. The opening turn asks only which prerequisites you have and what you want, in one go, and takes those prerequisites from the field's own textbook or review. After the motivation and vocabulary, each part of the field gets a turn of its own, and a part you got wrong stays marked shaky until you get it right. Say "derivation first" or "show the map" whenever you like. Jump to the ladder.

If it helps you, a ⭐ makes it easier for other people to find. Watch the repo if you want to hear when the rules change; the changelog has moved most weeks, mostly because I keep using it on my own reading and finding places where it breaks.

Quick start

Paste this into your agent, whichever one you use:

Install the field-onboarding skill from
https://github.com/ljx-chase/research-field-onboarding, following the
installation section of the repo's AGENTS.md.

It works out where its own skills directory is. If you would rather do it by hand, AGENTS.md has the per-agent paths, and on Claude Code there is a one-liner:

npx skills add ljx-chase/research-field-onboarding -g

Claude web and ChatGPT want a zip; there is one in releases.

That is the whole setup. Users never run the bundled state script or edit its JSON; capable agents use it internally, and other agents fall back to the same prompt workflow.

Then try:

I understand nonlinear optics but not topological photonics. Guide me into the field step by step.

A good first response names the prerequisites and asks which ones you already have. It does not open with a paragraph of definitions.

Why this exists

Ask an assistant about a field you do not know, and it will answer at the level of someone who already does. The answer is correct. You cannot use it, and you cannot tell which of its twelve terms you were supposed to already know. Ask again and you get more terms.

Without it

You: Explain attention in transformers.

Assistant: Transformers replace recurrence with self-attention, projecting each token into query, key and value vectors, scoring them by dot product, normalizing with softmax, then concatenating multiple heads through a residual connection into a feed-forward block with layer norm and positional encoding.

You: Which of those am I supposed to already know?

Assistant: Self-attention maps each token to three learned projections, then takes a similarity-weighted sum over the sequence.

With it

You: Explain attention in transformers.

Assistant: Before we start: the shortest bridge from what you already know. Mark each one.

• Matrix multiplication and dot products — attention is a stack of matrix operations
• Softmax and probability normalization — this is where the attention weights come from
• Embeddings and sequence representation — how text becomes vectors the model can work on

used it / learned it / new

And what is it for: a read a paper / b follow talks / c build one / d judge whether it fits your work? A reply like 1 used, 2 new, 3 learned, a is enough.

Twelve terms. It asks about three of them instead of throwing all twelve at you at once, then teaches from wherever you actually are, one step per turn.

The problem is not that the model knows too little. It is that nobody asked you what you know before choosing where to start.

This skill is not a knowledge base, and it does not carry a library of field summaries. A field-specific corpus goes stale the moment it is written, and shipping five fields well would silently degrade every other field. What it carries instead is a discipline:

locate the reader first, then teach upward from there, one step at a time.

The model already knows the physics. What it lacks, by default, is the habit of asking where you are standing before it starts pointing.

Pick how it explains

Same content, different order. Physical picture first unless you say otherwise; say "balanced" or "derivation first" at any time and the next turn switches.

• Physical picture first (default) — intuition, geometry, limiting cases and concrete phenomena first, then the equations with every term interpreted.
• Balanced — intuition and formalism advance together.
• Derivation first — definitions, assumptions and the mathematical steps first, physical interpretation after.

These are teaching priorities, not difficulty levels. Nothing is dumbed down in any of them: jargon gets explained, but equations, assumptions, scales and limitations stay.

These three styles govern the teaching ladder. For a register that governs every answer in ordinary research discussion, whether or not you are being onboarded into anything, see pick-your-professor.

The rules

Twelve rules. Full text in SKILL.md.

1. Name the prerequisites from the field's own structure; don't ask "what's your background".
2. Anything marked used it is an anchor and never gets explained again.
3. Separate style from level; plain language must keep the real science.
4. One step per turn: motivation, vocabulary, one part of the field at a time, frontier.
5. End each rung with a question the rung itself answers, not "make sense?".
6. Say where the analogy breaks, every time you use one.
7. State which sign, unit or normalization convention you are using.
8. Verified with a DOI, or labelled "from memory, unverified". No third option.
9. Say whether the field is settled before teaching it as if it were.
10. Answer short questions short. Offer the ladder once, then drop it.
11. Take the field's structure from a textbook, review or syllabus; don't invent it.
12. Name what was left out, and what each skipped part is for.

What it does differently

1. It names your gaps for you. Not "what's your background?" — you cannot audit a gap you cannot see. The agent looks up the field's structure, takes the three upstream frameworks it assumes, lists them with a one-clause gloss, and asks you to mark each as used it, learned it, or new. Then it uses the marks: anchors are never re-taught, black boxes are declared as black boxes, and a load-bearing gap gets built before anything stands on it.

2. It separates teaching style from technical level. Physical picture first by default; say "balanced" or "derivation first" to switch. The default treats you as a capable researcher entering a new field: jargon is unpacked, while mechanisms, assumptions, and equations stay. The opening questions fit in one control where the host has one, or one line of numbers where it does not.

3. It teaches one step at a time. Motivation and vocabulary once, then each part of the field in its own turn, then the frontier. Each step ends with a real diagnostic — a prediction, a restatement, a forced choice — not "does that make sense?", which always gets a yes. Wrong answers get a different explanation, not a louder one.

4. It will not invent a reference. Reading paths are where models fabricate, and a plausible title with a plausible year costs you an afternoon. Every named work is either verified in-session with a DOI or arXiv ID, or explicitly labelled from memory, unverified. There is no third option, and an identifier that was not retrieved is never attached.

5. It knows when to stay out of the way. Ask a narrow factual question and you get a narrow factual answer. The ladder is offered once, in one line, and dropped if you do not take it.

Who this is for

This is built around a specific situation: you are competent somewhere, and you need to get oriented somewhere else.

• Graduate students and postdocs reading outside their specialty, where the hard part is not intelligence but knowing which of the unfamiliar terms you are allowed to skip.
• Researchers evaluating a method from a neighbouring field, who need to decide whether it fits their problem rather than to master it.
• Anyone who has been burned by a fabricated citation, and wants a reading path where every item is labelled.
• Reviewers, examiners, and supervisors who have to parse a manuscript outside their area.

It is deliberately less useful if you are a specialist asking inside your own field, or if you want a one-line factual answer. In both c