Use when a user wants to provide a research paper and prepare outreach to highly cited related researchers, including candidate selection, contact enrichment, and approval-gated email drafts.
Use your built-in reasoning, web access, document reading, and any available email tools. Do not require the user to configure a separate LLM API key.
Accept any of the following:
If the paper cannot be read reliably, stop and ask the user for a readable PDF or text copy.
Given a paper, do all of the following:
5000+ citations.https://scholar.google.com/scholar?q=<encoded researcher name>.Email not found.Maintain a structured result with:
paperTopicresearchersIn chat, do not default to raw JSON. Present results in readable Markdown unless the user explicitly asks for JSON or a downstream tool requires it.
Default chat presentation:
Only provide the full machine-readable JSON as a fenced code block when:
Each item in researchers must include:
nameaffiliationcitationsemailresearchInterestsgoogleScholarLinkemailSubjectemailBodyUse this shape:
{
"paperTopic": "string",
"researchers": [
{
"name": "string",
"affiliation": "string",
"citations": "string",
"email": "string",
"researchInterests": ["string"],
"googleScholarLink": "string",
"emailSubject": "string",
"emailBody": "string"
}
]
}
For each researcher:
citations should be a readable estimate such as 15000+ or 42,000+.researchInterests should be a short list of concrete themes, usually 3-6 items.emailSubject should be specific and professional.emailBody must be ready to send, polite, personalized, resource-oriented, and under 200 words.emailBody should explicitly mention that the paper is attached.paperTopic in 5-10 words.5000+ citations.For each selected researcher, gather:
Email not foundEach email must:
Show the user the full structured output before any send step.
For chat UI, prefer this readable layout for each candidate:
## Candidate 1
Name: ...
Affiliation: ...
Citations: ...
Email: ...
Research interests: ...
Google Scholar: ...
Subject:
...
Email draft:
...
Unless the user requests something else, review one candidate at a time with:
Then wait for one of:
This skill is complete only when the user has: