A Claude skill that takes a viral post, screenshot, meme, headline, or forwarded message and splits it into its parts the way a prism splits light: the underlying story on one side, the framing and technique on the other.
You paste the thing that’s going around. It reconstructs what’s actually known and shows you how the post is built. Then it hands the judgment back to you.
Why this exists
Viral content wins by compression. It takes a complicated situation and strips it down to one emotional, shareable claim, and the nuance is exactly what gets thrown away in the stripping: the missing denominator, the recycled photo, the real grievance underneath the spin, the half of the story that doesn’t fit the frame. By the time a post is doing numbers, it has usually been flattened past the point where it tells you anything true about the world.
The reflex answer to this is the fact-check: a verdict, a rating, a “debunked” stamp. But a verdict is itself a compression. It replaces one thing to repeat uncritically with another, and it asks you to outsource the judgment to whoever issued the ruling. It also tends to pick a side and defend it, which means it goes quiet exactly when the “respectable” counter-narrative is spinning too.
This skill does the opposite of compressing. It re-inflates. The thing it is against is flattening, not falsehood.
What it does
Paste a post and it works in four moves:
- Reconstructs the underlying story. It searches the web, prefers primary and contemporaneous reporting over aggregators and other viral posts, and cites its sources inline so you can follow them.
- Names the techniques precisely. It quotes the exact phrases doing the work and names the specific move (loaded language, missing base rate, decontextualized image, manufactured consensus, and so on) rather than gesturing at “bias.” It carries a catalog of about thirty such techniques as a lens.
- Applies nuance in both directions. It names the post’s legitimate kernel, the real thing it’s wrapped around, and then turns the same lens on the counter-take and on the reader’s own likely side.
- Leaves the unsettled parts unsettled. If the sources don’t resolve something, it says so and stops, rather than forcing a conclusion.
What it won’t do
It will not tell you whether the post is true or false. No verdict, no rating, no score, no bottom-line ruling of any kind. It will not tell you what to believe, how to feel, or how to vote. If you want a thumbs-up or thumbs-down, this is the wrong instrument.
It also won’t manufacture spin that isn’t there. Sometimes a post is a fair, proportionate representation of its subject, and when that is the case, it says so. Inventing balance is its own kind of distortion.
Who it’s for
Anyone who wants to think clearly about something going around, and who would rather be handed a clearer view than a ruling to repeat. It assumes you are capable of judging for yourself and is built to return that judgment to you instead of keeping it.
How to use it
- Upload the skill through claude.ai → Settings → Capabilities → Skills. There is nothing to configure.
- Enable web search in your Claude settings; the skill leans on it to reconstruct the story and will tell you plainly when it couldn’t verify the facts.
- In any conversation, paste the post or upload a screenshot of it. The skill should trigger on its own; if not, ask Claude to “use the nuance-finder skill.”
The layperson README in the bundle has a step-by-step install for non-technical users.
Status
Working draft. Built for personal use and shared as-is. It is only as good as what the web will tell it: for fast-moving or genuinely obscure stories it may not reconstruct much, and it is built to say so rather than guess. Treat its citations the way you’d treat any research, by following the links and checking them yourself.
Issues, forks, and pull requests welcome.