Research synthesis

Skill SkillMedev/skills/skills/research-synthesis

Turns many sources into a decision-ready synthesis - atomic claims clustered into named themes, surfaced tensions between sources, gaps no source answers, and explicit so-what/now-what implications with confidence ratings. Use when someone says "synthesize this research", "what do all these reports agree on", "pull the themes out of these articles", or has a pile of sources and needs conclusions rather than source-by-source summaries. Do NOT use for synthesizing user interview transcripts - use interview-synthesis instead; for a formal academic literature review, use literature-review.From its SKILL.md

Install

npx -y skills add SkillMedev/skills --skill research-synthesis

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.5 KB, 684 tokens by cl100k_base, as published. Nobody here has run it

Research Synthesis

Synthesis is not summary. A summary repeats what each source said; a synthesis finds the patterns across sources and says what it means. This skill turns 20+ sources into a decision-ready document.

Step 1: Capture, don't just read

As you go through each source, extract atomic notes (one idea per note):

Source: <author, title, date>
Claim: <a single finding or argument, in your own words>
Evidence: <what backs it - data, study, anecdote>
Confidence: <strong / moderate / weak>

Keep claims atomic so you can later regroup them across sources. Tag each with a topic.

Step 2: Cluster into themes

Lay all the claims out and group them by what they're about, regardless of source. Patterns emerge when ideas from different authors sit together. Name each cluster with a theme - a sentence that captures the shared insight, not just a topic word.

Theme: "Remote teams ship faster but report weaker belonging."
  - [Source A] productivity up 12% remote
  - [Source C] async cuts cycle time
  - [Source B] belonging scores down in fully-remote orgs

Step 3: Find tensions and disagreements

The most valuable output is where sources conflict. For each tension:

State both positions and who holds them.
Hypothesize why they differ (different populations? definitions? time periods? incentives?).
Note what evidence would resolve it.

Don't paper over disagreement - surfacing it is the point.

Step 4: Identify gaps

What's missing? Note:

Questions no source answers.
Weakly-supported claims everyone repeats (an echo, not evidence).
Populations / contexts not studied.

Gaps tell the reader where to be cautious and where new research/work is needed.

Step 5: Write the "so what"

End every theme and the document overall with implications:

So what? - why does this matter for the decision at hand?
Now what? - what should we do given this? Concrete recommendations.

A synthesis that doesn't reach "so what" is just organized note-taking.

Output structure

# Synthesis: <question>

## Executive summary (the so-what, up front)
## Themes
  ### Theme 1 - <insight sentence>
    Evidence: [sources]. So what: <...>
## Tensions & disagreements
## Gaps & open questions
## Implications & recommendations
## Source list with confidence ratings

Tips

Synthesize across, not down - organize by theme, never source-by-source.
Weight by quality - three weak blog posts don't outweigh one rigorous study.
Quote sparingly; paraphrase into your shared framework.
Show your confidence so readers can calibrate.

Anti-patterns

A list of source summaries with a conclusion bolted on - that's not synthesis.
Ignoring disagreement to present a tidy narrative.
Treating volume of mentions as truth - popularity isn't evidence.
Stopping at "what" without "so what / now what".

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.