Set it up once, then steal these. Each play is a prompt you paste straight into your AI — for backing up a claim, building a reading list, finding the paper the field is built on, or grabbing something you can read tonight.
You've written a claim and your tutor wants a credible source behind it.
I wrote: “Hawker centres strengthen social ties across income groups.” Find me a credible academic source that supports this, with a link I can open.
Under the hood: Runs search_ntu_library on the claim's keywords, reads the abstracts, and returns a source that actually backs the point — plus how to access it.
Or call the skill: /cite_a_claim claim: “…”New essay, blank page. You want 5–6 solid sources to start from.
Build me a reading list of 6 sources for an essay on the future of Singapore's hawker culture. Mix open-access and library items, and tell me which I can download now.
Under the hood: Searches broadly, groups by access (download-now first), and summarises what each source argues so you can pick fast.
Or call the skill: /reading_list topic: “…”You want the seminal paper the whole field cites — to anchor your argument.
What are the most-cited academic works on urban food heritage? Then check which ones NTU can give me access to.
Under the hood: Runs find_influential_research (Crossref) to rank by citations, then cross-checks the top titles through search_ntu_library for NTU access.
Or call the skill: /most_cited topic: “…”Deadline's tomorrow. You need sources you can open right now, not request.
Find recent peer-reviewed articles on climate adaptation in cities that I can download or read online tonight — skip anything print-only.
Under the hood: Searches with resource_type=articles, then filters to items tagged open-access or online-via-NTU-login and hands you the links.
Or call the skill: /downloadable_only topic: “…”You read a statistic somewhere and need the original, citable source.
I saw a claim that street food vending supports a large share of urban employment in Southeast Asia. Find the original research behind a figure like that and give me the reference.
Under the hood: Searches the topic, reads abstracts to find the study reporting the figure, and returns the primary source with its DOI.
Or call the skill: /trace_a_fact claim: “…”A reading references a specific book and you want to know if NTU has it.
Does NTU have the book ‘The Hawker Chan Story’? If so, tell me whether it's online or on a shelf, and where.
Under the hood: Runs search_ntu_library with resource_type=books and reports the access tag and location.
Or call the skill: /find_a_book title: “…”A topic returns thousands of hits; a claim returns the right few. Give it your actual argument, not just the theme.
Find sources that argue hawker centres reduce social isolation — not just about hawker food generally.Ask for downloadable-only when you're in a hurry. The tool tags each source, so it can filter to what you can actually open.
Only show me ones I can download for free or open with my NTU login.Ask it to summarise what each source argues and whether it fits your point — don't accept a bare list of titles.
For each result, summarise the argument in one line and say if it supports my thesis.Find the most-cited work across the field, then check what NTU can actually give you. Influence first, access second.
Most-cited papers on this topic, then which ones NTU has access to.Once you've picked a source, get a clean citation with the DOI or permalink so you can drop it straight into your essay.
Give me an APA citation for the second source, with the DOI.Tell it books vs. articles, and “recent” when currency matters. It maps that to the right filters.
Just journal articles from the last five years, please.Each play above has a matching skill — an MCP prompt your client shows as a slash command. You don't type the whole instruction; you pick the skill and give it your content.
Run the skill and pass your content as the argument — the client fills the rest.
/cite_a_claim claim: "Hawker centres strengthen social ties."Prefer plain language? Skip the slash command — the assistant already knows the tools and rules.
Find me a source that backs this sentence, with a link.Find a credible source that backs a specific claim you've written, with a link.
Build a starter reading list of sources on a topic, marked by how to get each.
Find the most-cited work on a topic, then check NTU access.
Find sources on a topic you can read online or download right now.
Trace a statistic or fact back to its original, citable source.
Check whether NTU has a specific book and where to get it.
You don't have to teach the AI how to behave — the server sends its guidance as MCP instructions the moment your client connects. That's why it picks the right tool and leads with sources you can actually get.
The golden rule, baked in: it always reads the abstract before recommending a source — and summarises what each one argues, so you never cite something on its title alone.
No files to edit. In Claude Desktop, claude.ai, or the mobile app:
Works on Free, Pro, Max, Team & Enterprise (Free = 1 custom connector).
In ChatGPT (Plus, Pro, Business, Enterprise or Edu — not Free):
Custom connectors aren't on the Free plan. Business/Enterprise need an admin to turn on Developer mode.
Use the Gemini CLI — one command adds it (no manual file edits):
gemini mcp add -t http ntu-library https://ntu-library.openskillshub.org/mcp
On Gemini Enterprise, an admin connects a custom MCP data store (see the Enterprise link). The consumer Gemini app doesn't take custom MCP servers yet.
One command in your terminal:
claude mcp add --transport http ntu-library https://ntu-library.openskillshub.org/mcp
Add to ~/.cursor/mcp.json (or any MCP client that speaks Streamable HTTP):
{
"mcpServers": {
"ntu-library": {
"url": "https://ntu-library.openskillshub.org/mcp"
}
}
}