AGENT SKILL

research-summarizer

Structured research summarization agent skill for non-dev users. Handles academic papers, web articles, reports, and documentation. Extracts key findings, generates comparative analyses, and produces properly formatted citations. Use when: user wants to summarize a research paper, compare multiple sources, extract citations from documents, or create structured research briefs. Plugin for Claude Code, Codex, Gemini CLI, and OpenClaw.

License 许可
MIT
最近更新
2026-08-16
Stars
★ 0
Forks
0

01 前言 Front Matter

name
research-summarizer
license
MIT
metadata
{ "author": "Alireza Rezvani", "updated": 1773619200000, "version": "1.0.0", "category": "product" }
description
Structured research summarization agent skill for non-dev users. Handles academic papers, web articles, reports, and documentation. Extracts key findings, generates comparative analyses, and produces properly formatted citations. Use when: user wants to summarize a research paper, compare multiple sources, extract citations from documents, or create structured research briefs. Plugin for Claude Code, Codex, Gemini CLI, and OpenClaw.

02 资源文件 4

03 SKILL.md 正文

Research Summarizer

Read less. Understand more. Cite correctly.

Structured research summarization workflow that turns dense source material into actionable briefs. Built for product managers, analysts, founders, and anyone who reads more than they should have to.

Not a generic "summarize this" — a repeatable framework that extracts what matters, compares across sources, and formats citations properly.


Scope — Distinct From the research/ Domain

This skill summarizes documents the user already has (papers, articles, reports pasted or attached). It performs no web search and needs no MCP server. It is NOT:

  • research/litreview — academic literature discovery and review-guide generation (finds papers via Consensus/academic APIs)
  • research/dossier — entity due-diligence built from live web research
  • research/notebooklm — drives Google's NotebookLM product UI
  • research/research — the router for open-ended "research [topic]" requests that require searching

If the user asks you to find sources rather than digest supplied ones, route to the research/ domain instead.


When This Skill Activates

Recognize these patterns from the user:

  • "Summarize this paper / article / report"
  • "What are the key findings in this document?"
  • "Compare these sources"
  • "Extract citations from this PDF"
  • "Give me a research brief on [topic]"
  • "Break down this whitepaper"
  • Any request involving: summarize, research brief, literature review, citation, source comparison

If the user has a document and wants structured understanding → this skill applies.


Workflow

Workflow 1 — Single Source Summary

  1. Identify source type

    • Academic paper → use IMRAD structure (Introduction, Methods, Results, Analysis, Discussion)
    • Web article → use claim-evidence-implication structure
    • Technical report → use executive summary structure
    • Documentation → use reference summary structure
  2. Scaffold the brief — python3 scripts/format_summary.py --template academic (or article/report/executive per source type), then fill in every section from the source:

    Title: [exact title]
    Author(s): [names]
    Date: [publication date]
    Source Type: [paper | article | report | documentation]
    
    ## Key Thesis
    [1-2 sentences: the central argument or finding]
    
    ## Key Findings
    1. [Finding with supporting evidence]
    2. [Finding with supporting evidence]
    3. [Finding with supporting evidence]
    
    ## Methodology
    [How they arrived at these findings — data sources, sample size, approach]
    
    ## Limitations
    - [What the source doesn't cover or gets wrong]
    
    ## Actionable Takeaways
    - [What to do with this information]
    
    ## Notable Quotes
    > "[Direct quote]" (p. X)
    
  3. Assess quality

    • Source credibility (peer-reviewed, reputable outlet, primary vs secondary)
    • Evidence strength (data-backed, anecdotal, theoretical)
    • Recency (when published, still relevant?)
    • Bias indicators (funding source, author affiliation, methodology gaps)

Workflow 2 — Multi-Source Comparison

  1. Collect sources (2-5 documents)

  2. Summarize each using the single-source workflow above

  3. Build comparison matrix

    | Dimension        | Source A        | Source B        | Source C        |
    |------------------|-----------------|-----------------|-----------------|
    | Central Thesis   | ...             | ...             | ...             |
    | Methodology      | ...             | ...             | ...             |
    | Key Finding      | ...             | ...             | ...             |
    | Sample/Scope     | ...             | ...             | ...             |
    | Credibility      | High/Med/Low    | High/Med/Low    | High/Med/Low    |
    
  4. Synthesize

    • Where do sources agree? (convergent findings = stronger signal)
    • Where do they disagree? (divergent findings = needs investigation)
    • What gaps exist across all sources?
    • What's the weight of evidence for each position?
  5. Produce synthesis brief

    ## Consensus Findings
    [What most sources agree on]
    
    ## Contested Points
    [Where sources disagree, with strongest evidence for each side]
    
    ## Gaps
    [What none of the sources address]
    
    ## Recommendation
    [Based on weight of evidence, what should the reader believe/do?]
    

Workflow 3 — Citation Extraction

  1. Run the extractor — python3 scripts/extract_citations.py document.txt --output json detects DOI/URL/author-year/numbered citations and deduplicates them
  2. Review and format the extracted list in the requested style (APA 7 default); manually catch citations the regex missed
  3. Classify citations by type:
    • Primary sources (original research, data)
    • Secondary sources (reviews, meta-analyses, commentary)
    • Tertiary sources (textbooks, encyclopedias)
  4. Output sorted bibliography with classification tags

Supported citation formats:

  • APA 7 (default) — social sciences, business
  • IEEE — engineering, computer science
  • Chicago — humanities, history
  • Harvard — general academic
  • MLA 9 — arts, humanities

Tooling

scripts/extract_citations.py

CLI utility for extracting and formatting citations from text.

Features:

  • Regex-based citation detection (DOI, URL, author-year, numbered references)
  • Multiple output formats (APA, IEEE, Chicago, Harvard, MLA)
  • JSON export for integration with reference managers
  • Deduplication of repeated citations

Usage:

# Extract citations from a file (APA format, default)
python3 scripts/extract_citations.py document.txt

# Specify format
python3 scripts/extract_citations.py document.txt --format ieee

# JSON output
python3 scripts/extract_citations.py document.txt --format apa --output json

# From stdin
cat paper.txt | python3 scripts/extract_citations.py --stdin

scripts/format_summary.py

CLI utility that emits blank structured summary scaffolds — you (the model) fill them in from the source. It does not analyze content itself.

Features:

  • 6 templates: academic, article, report, executive, comparison, literature
  • Configurable scaffold depth (brief, standard, detailed)
  • Text and JSON output for downstream tooling

Usage:

# Generate structured summary template
python3 scripts/format_summary.py --template academic

# Brief executive summary format
python3 scripts/format_summary.py --template executive --length brief

# All templates listed
python3 scripts/format_summary.py --list-templates

# JSON output
python3 scripts/format_summary.py --template article --output json

Quality Assessment Framework

Rate every source on four dimensions:

Dimension High Medium Low
Credibility Peer-reviewed, established author Reputable outlet, known author Blog, unknown author, no review
Evidence Large sample, rigorous method Moderate data, sound approach Anecdotal, no data, opinion
Recency Published within 2 years 2-5 years old 5+ years, may be outdated
Objectivity No conflicts, balanced view Minor affiliations disclosed Funded by interested party, one-sided

Overall Rating:

  • 4 Highs = Strong source — cite with confidence
  • 2+ Mediums = Adequate source — cite with caveats
  • 2+ Lows = Weak source — verify independently before citing

Summary Templates

See references/summary-templates.md for:

  • Academic paper summary template (IMRAD)
  • Web article summary template (claim-evidence-implication)
  • Technical report template (executive summary)
  • Comparative analysis template (matrix + synthesis)
  • Literature review template (thematic organization)

See references/citation-formats.md for:

  • APA 7 formatting rules and examples
  • IEEE formatting rules and examples
  • Chicago, Harvard, MLA quick reference

Proactive Triggers

Flag these without being asked:

  • Source has no date → Note it. Undated sources lose credibility points.
  • Source contradicts other sources → Highlight the contradiction explicitly. Don't paper over disagreements.
  • Source is behind a paywall → Note limited access. Suggest alternatives if known.
  • User provides only one source for a compare → Ask for at least one more. Comparison needs 2+.
  • Citations are incomplete → Flag missing fields (year, author, title). Don't invent metadata.
  • Source is 5+ years old in a fast-moving field → Warn about potential obsolescence.

Installation

One-liner (any tool)

git clone https://github.com/alirezarezvani/claude-skills.git
cp -r claude-skills/product-team/research-summarizer ~/.claude/skills/

Multi-tool install (run from the claude-skills repo root)

./scripts/convert.sh --skill research-summarizer --tool codex|gemini|cursor|windsurf|openclaw

OpenClaw

clawhub install cs-research-summarizer

Verification Loop

Before delivering any brief, check:

  1. Every Key Finding cites a location in the source (section, page, or quote) — no unanchored claims.
  2. python3 scripts/extract_citations.py <file> --output json exits 0 and its total matches the bibliography count in your output (investigate any gap).
  3. Each source carries a 4-dimension quality rating (table above); weak sources are flagged, not silently included.
  4. For comparisons: the matrix has one row per dimension and one column per source — no source skipped.
  5. Nothing was invented: missing metadata is marked "not stated", never filled in.

Related Skills

  • product-analytics — Quantitative analysis. Complementary — use research-summarizer for qualitative sources, product-analytics for metrics.
  • competitive-teardown — Competitive research. Complementary — use research-summarizer for individual source analysis, competitive-teardown for market landscape.
  • content-production — Content writing. Research-summarizer feeds content-production — summarize sources first, then write.
  • product-discovery — Discovery frameworks. Complementary — research-summarizer for desk research, product-discovery for user research.

04 同仓库更多 Skills

GongYuanCaiJi
GongYuanCaiJi/dsh-claude-skills

syllabus

Generates a curated supplementary reading list from any course syllabus using Consensus academic search. Grill-me intake (syllabus input format + course audience + year range) plus a grouping forcing-options checkpoint before any search runs — so the reading list matches the course's level and recency need. Parses the syllabus to extract topics and learning outcomes, searches Consensus for recent peer-reviewed papers per topic, and produces a professionally formatted .docx with clickable Consensus links, plain-language summaries calibrated to audience level, and Bloom-higher-order discussion questions tied to course learning goals. Use when the user uploads a syllabus, course outline, or curriculum document and wants supplementary readings (e.g., 'create a reading list from this syllabus', 'find recent papers for my course') — even casual mentions with a syllabus attached should trigger this skill.

★ 0
GongYuanCaiJi
GongYuanCaiJi/dsh-claude-skills

research

Default entry point for any research request — a hybrid router that classifies the question deterministically and either delegates to a specialist research skill (pulse for trends/sentiment, grants for NIH funding, litreview for academic literature, syllabus for course reading, patent for prior-art + IP landscape, dossier for entity research) or runs its own plan-decompose-multi-source-search-synthesize-cite fallback workflow when no specialist matches. Always surfaces the routing decision so users can override. Use when the user makes any research request that doesn't obviously match a more-specific specialist skill (e.g., "research [topic]", "look into [topic]", "what do we know about [topic]", "investigate [topic]", "find me information on [topic]", "do some research on [topic]", "I need to understand [topic]"). Output is a markdown briefing (default) or .docx document (on request) with full citations and an audit log.

★ 0
GongYuanCaiJi
GongYuanCaiJi/dsh-claude-skills

pulse

Multi-source recency research skill that takes the pulse of any topic across Reddit, Hacker News, the open web, and optionally X/Twitter within a configurable recent window (default 30 days). Forcing intake clarifies topic specificity, angle (trend/sentiment/problems/opportunities/comparison), time window, and platform scope before searching. Returns a synthesized briefing with citations, engagement metrics, and cross-platform pattern analysis. Use when the user requests multi-source recency intelligence on a topic (e.g., 'pulse on [topic]', 'what's happening with [topic]', 'what are people saying about [topic]', 'current conversation about [topic]', 'take the pulse of [topic]', 'trending: [topic]', 'find me info on [topic]'), and for competitor research, trend discovery, tool comparisons, and audience sentiment analysis.

★ 0
GongYuanCaiJi
GongYuanCaiJi/dsh-claude-skills

patent

Patent prior-art and landscape intelligence skill — not generic patent help. Commits to one of five sub-use-cases via forcing intake (novelty search / freedom-to-operate / competitive landscape / acquisition diligence / litigation prior-art) before any search runs. Searches Google Patents, Espacenet, USPTO, and optionally Lens.org for citation-graph signals. Output is an editable Word document (.docx) with verdict, ranked closest art (claim-text extracted), CPC-class-aware landscape, family-resolved hits, geographic coverage, FTO flags where applicable, strategy recommendations, and full audit log. Use when the user asks for patent searching or analysis (e.g., 'prior art search for [invention]', 'freedom to operate analysis for [product]'). Produces search signal, not legal advice — always recommends consulting a patent attorney before filing or licensing decisions. Trademark, copyright, and trade-secret questions are out of scope.

★ 0
GongYuanCaiJi
GongYuanCaiJi/dsh-claude-skills

notebooklm

Browser automation skill for controlling Google's NotebookLM. Use when the user wants anything done in NotebookLM (e.g., 'open NotebookLM', 'check my [name] notebook', 'ask my notebook about X', 'add [source] to NotebookLM', 'generate a Video Overview from my notebook', 'use NotebookLM Studio'). Handles reading and querying notebooks, adding sources (URLs, text, files, YouTube links, synthesized content), generating Studio outputs (Audio/Video Overviews, Mind Maps, Reports incl. Briefing Doc/Study Guide/FAQ, Flashcards, Quiz, slide decks, infographics — discover the exact set from the live Studio panel; the UI evolves fast), and creating new notebooks. Requires browser automation environment — fails gracefully when unavailable.

★ 0
GongYuanCaiJi
GongYuanCaiJi/dsh-claude-skills

litreview

Academic literature orientation skill that searches papers via free keyless APIs (PubMed E-utilities + OpenAlex) by default — with the Consensus MCP as an optional enhancement lane when connected — builds a strategic search plan using PICO (default) or SPIDER / Decomposition / hybrid as fallbacks, and synthesizes findings into a formatted Word (.docx) research guide. Grill-me intake (research question specificity + framework hint + tentative depth) before the recon search; a second forcing checkpoint after Phase 2 confirms framework + sub-areas + depth before searches consume budget. Configurable depth (5/10/20 queries) controls coverage vs. speed. Output is a 'launching pad' — an orientation guide that lets a researcher dive in confidently, not a finished review. Use when the user starts literature-oriented research (e.g., 'litreview on [topic]', 'literature review on [topic]', 'I'm starting a literature review on X', 'I'm writing a paper on X', 'help me research X', 'I'm doing research on X', 'can you help me research X'). Do NOT use for single one-off paper searches wanting a quick list — that's a plain PubMed/OpenAlex (or Consensus) query.

★ 0