Help us improve
Share bugs, ideas, or general feedback.
Checks every statistic in a draft for misleading figures, incorrect comparisons, and missing sources. Useful for journalists and writers fact-checking data before publication.
npx claudepluginhub ur-grue/autopunk-media-skills --plugin autopunk-media-skillsHow this skill is triggered — by the user, by Claude, or both
Slash command
/autopunk-media-skills:statistics-verifierThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Runs a structured checklist over every statistic in your draft — flagging figures that look wrong, comparisons that may mislead, and numbers that need a primary source before they are safe to publish.
Analyzes statistics in draft articles for misrepresentation, missing context, misleading framing, and sourcing gaps before publication.
Detects statistical errors, logical fallacies, and methodological issues in research content. Use for validating statistics, auditing quantitative claims, or checking methodology.
Extracts factual claims from PR copy, verifies each claim independently with concrete citations, and flags unresolved risk before journalist-facing drafts are sent.
Share bugs, ideas, or general feedback.
Runs a structured checklist over every statistic in your draft — flagging figures that look wrong, comparisons that may mislead, and numbers that need a primary source before they are safe to publish.
Required: The full draft text, or at minimum every sentence containing a statistic. For each figure, provide the source you used — paste the relevant excerpt or describe it (e.g. "government report, page 12: 'unemployment rose to 7.4%'").
Optional: The type of story (news report, feature, documentary script, social media post). Any statistics you already know are contested. The intended audience, if it affects how precise the language needs to be.
A numbered list of every statistic found, each followed by: the flag level, the specific issue (if any), and the recommended action. After the list, a one-paragraph summary: overall picture of the draft's statistical health, the highest-priority fix, and a confidence rating (Low / Medium / High) for the draft as a whole. Total length scales with the number of statistics — roughly 100 words per flagged figure. Plain language throughout; no statistical jargon without immediate explanation.
Story type: News report for a regional news website
Draft excerpt: "Youth unemployment in the region has doubled over the past two years, reaching 18% — the worst rate in the country. Local officials say the figure is misleading because many young people are in part-time work or training schemes. Meanwhile, the national rate stands at 9%, meaning young people here are twice as likely to be out of work as their peers elsewhere. The regional council has spent £4.2 million on job creation programmes since 2021, though critics say that amounts to less than £200 per unemployed young person."
Sources provided:
1. "Youth unemployment has doubled over the past two years" Flag: CHECK BEFORE PUBLISHING The regional statistics office gives 9.1% two years ago and 18.3% now. That is a near-doubling (ratio: 2.01x) so "doubled" is defensible, but only just. The report likely refers to Q3-to-Q3 — confirm the comparison uses the same quarter in both years to avoid seasonal distortion. Recommended action: confirm the exact comparison periods with the statistics office and consider changing "doubled" to "nearly doubled" or "risen by 101%" to be precise.
2. "Reaching 18%" Flag: PUBLISH (with minor caveat) The regional statistics office reports 18.3%, which rounds correctly to 18%. No issue with the figure itself, but note your source says Q3 of last year — if the report is recent, flag this as last year's data in the story, not current.
3. "The worst rate in the country" Flag: DO NOT PUBLISH AS WRITTEN No source has been provided for this claim. A ranking claim requires a complete dataset of all regions or a direct statement from the national statistics authority. A single region's figure cannot itself establish a national ranking. Recommended action: check whether the national statistics authority publishes a regional breakdown table. If no such table exists or the region does not rank first, remove or soften the claim ("one of the highest rates," "among the worst-performing regions").
4. "The national rate stands at 9%, meaning young people here are twice as likely to be out of work" Flag: CHECK BEFORE PUBLISHING Two issues. First, the national figure is 9.2%, not 9% — a minor but correctable inaccuracy. Second, "twice as likely" is mathematically correct (18.3% vs 9.2% ≈ 1.99x) but only if both figures use the same definition of "youth" and "unemployed." Confirm both use the ILO unemployment definition and the same 16–24 age band — if the regional figure uses a different age bracket, the comparison breaks. Recommended action: check methodology footnotes in both reports for age band and definition consistency; correct "9%" to "9.2%."
5. "£4.2 million on job creation programmes since 2021" Flag: PUBLISH Matches the council press release. No issue. Note this is committed funding, not necessarily spent — if the distinction matters for your story, check whether the full amount has been disbursed.
6. "Less than £200 per unemployed young person" Flag: DO NOT PUBLISH AS WRITTEN No source provided. This figure requires knowing the total number of unemployed young people in the region at a specific point in time — it is a derived calculation, not a figure that appears in any of your sources. If the calculation is wrong, it is a significant error. Recommended action: obtain the absolute number of unemployed 16–24-year-olds from the regional statistics office and perform the calculation explicitly: £4,200,000 divided by that figure. Show the working in your notes.
Summary: This draft has two statistics that must not publish as written (the national ranking claim and the per-person spending figure) and two that need source confirmation before publishing (the doubling claim and the national comparison). Two figures are clear. Overall confidence in the statistical layer of this draft as submitted: Low. Resolve the ranking claim and the per-person calculation first — both carry the highest risk of factual challenge after publication.