Theme quality auditor

A Claude skill that checks whether your synthesis themes are real findings or just topic labels, and asks a question to help you turn each label into a finding.

SynthesisOctober 12, 2026

Download the skill

Enter your details to download the skill file.

What it does

People label a synthesis theme "Onboarding" or "Pricing". That's a topic, not a finding. It says what people talked about, not what you learned from them.

You give it your list of themes from analysis, and your discussion guide if you have it. It tells you which themes are just topic labels, and for each one a question to help you turn it into a real finding, like "People give up on setup before they see any value". It never rewrites a theme for you. Use it before you share a synthesis or readout.

What you get back

  • Each theme marked as fine, worth a review, or reading as a topic summary, with the reason.
  • A question for each flagged theme that fits what was flagged: no claim, mirrors the guide, or thin evidence.
  • Themes that copy an item from your discussion guide, named against the guide item they mirror.
  • Thin-evidence notes kept apart from the topic verdict, so one strong participant isn't treated as a defect.
  • Description lines that summarise a domain ("Participants talked about...") picked out for a rethink.
  • A check on emergence language, with "developed" or "constructed" suggested in its place.

Install it in Claude, ChatGPT, Copilot or Gemini

The same skill file works in all four. Pick your tool for the steps.

Install in Claude

  1. 1

    Download the skill

    Fill in the short form to get the .skill file.

  2. 2

    Add it to Claude

    In the Claude app go to Settings → Capabilities → Skills and upload the file. In Claude Code, run the npx command shown after the form, or drop the folder into your project's skills folder.

  3. 3

    Just ask

    Ask in plain language. Claude picks up the skill on its own when the request matches.

How to use it

Once it's installed, ask in plain language. Your AI follows the skill's method when the request matches.

Example prompt 1
Check these themes before the readout. Which ones are just topics?
Example prompt 2
Here are our affinity groups and the discussion guide. Did any of the groups turn into the guide's own questions?
Example prompt 3
Audit this synthesis for topic summaries, and don't rewrite anything, just ask me the questions.

What it won't do

It won't write a theme for you. Deciding what a pattern means is the analyst's job, so the skill asks a question and leaves the claim to you. It also can't judge whether a claim is true to the data. It checks the shape of a theme, not the quality of the interpretation behind it.

Questions

Which kinds of thematic analysis does it support?

Reflexive thematic analysis, codebook approaches such as framework analysis, and affinity diagrams. In affinity mode it checks group labels the same way, since Contextual Design writes labels in the user's voice as statements rather than categories.

Will it flag a theme that only one participant supports?

It reports thin evidence as a separate note and never counts it toward a topic-summary verdict. Something said once can still be central, so the question it asks is whether anyone else said something that means the same.

What format do my themes need to be in?

Markdown with one heading per theme, quotes in blockquotes with the participant after them, or a JSON list of themes. The discussion guide is optional, one item per line.

More skills

See all skills