Building a Personal Research Assistant with Claude

There's a particular kind of cognitive overload that knowledge workers experience but rarely talk about: you have too much to read, most of it is important, and the act of reading it is taking so long that some of it is stale by the time you get to it. Industry reports, research papers, competitor blogs, long-form journalism, internal documents. The pile never gets smaller. It just changes shape.

The standard advice is to be more selective. Read less. Curate harder. And that's genuinely good advice, except it doesn't solve the underlying problem: your job may actually require processing a lot of information, and there's no version of "just read less" that works without something falling through the gaps.

A Claude-based research assistant gives you a different lever. Instead of reading less, you process faster. Not by skimming and missing things, but by using a skill to extract what matters so your reading time goes toward understanding and connecting, not just ingesting.

Designing the research assistant skill

The most important design decision is: what does "research" mean for you specifically? A management consultant's research needs are completely different from a product manager's, which are completely different from a journalist's. The skill works best when it's tuned to your domain and your specific extraction priorities.

You are a research assistant helping extract insights from documents.

Context about me and my work: {USER_CONTEXT}
Research question or purpose: {RESEARCH_PURPOSE}

Document to process: {DOCUMENT_TEXT}

Produce the following:

DOCUMENT SUMMARY (3-5 sentences)
What is this document about and what is its central argument or finding?

KEY INSIGHTS (5-8 bullet points)
The most important specific claims, data points, or findings.
Each point should be specific enough to be actionable or citable.

RELEVANCE TO MY WORK
Given my context above, what in this document is most relevant to me?

OPEN QUESTIONS
What questions does this document raise that it does not answer?

QUOTABLE PASSAGES
2-3 direct quotes worth preserving, with context.

CREDIBILITY NOTES
Any caveats about methodology, funding sources, or sample size.

Processing PDFs and articles in practice

The workflow is simple: copy the text content from a PDF or article, paste it into the skill with your context and research purpose, and review the output. For PDFs with complex formatting, the copied text can be messy. Tables often don't copy cleanly, and multi-column layouts can jumble the reading order. Spending thirty seconds cleaning up the paste before running the skill is worth it. You'll get much better output for the small extra effort.

The research purpose field is doing more work than it seems. "I'm reading this for background" produces very different output than "I'm evaluating whether this methodology is sound enough to cite in a client report." Be specific about why you're reading something.

Building toward synthesis

Individual document processing is useful. But the real power of a research assistant skill comes from synthesis: taking the extracted insights from five or ten documents and finding the patterns, contradictions, and gaps across them. After processing several documents, paste the key insights sections from each one into a separate synthesis prompt:

I have researched [TOPIC] across multiple sources.
Here are the key findings from each:

SOURCE 1 ([title/author]): {INSIGHTS}
SOURCE 2 ([title/author]): {INSIGHTS}

Please synthesize these findings:
1. What do they agree on?
2. Where do they contradict each other?
3. What claims appear in multiple sources?
4. What seems underexplored or missing?
5. What is my most defensible conclusion given this evidence?

This synthesis step is where the research assistant skill goes from "useful reading aid" to "genuine thinking partner." You're not just processing information faster; you're building a more complete picture than any individual source could give you.

Curious what other people have built in this space? Browse research and productivity skills on SkillsLib.ai, particularly useful if you're in a domain like law, medicine, finance, or academia where research has specific conventions the skill should know about.