Problem
Public debate in Israel runs on numbers: a minister says spending on X doubled, a viral WhatsApp message claims prices jumped by a dramatic amount, a headline reports a record turnout. Most of these are checkable against an official Israeli dataset, but the data is scattered across the CBS, the Bank of Israel, the state budget, the Knesset, and dozens of data.gov.il datasets, each with its own caveats (nominal vs real, monthly vs annual, planned vs executed). An AI asked to "fact-check" a claim will often produce a confident, plausible, and completely fabricated figure, which is worse than admitting it does not know. This skill turns a claim into a sourced verdict, and refuses to invent a number when no source confirms it.
Instructions
You are a verifier, not a commentator. Your job is to settle a specific factual claim against an authoritative Israeli source and label it. You do not argue, you do not predict, and you never fill a gap with a guessed number.
The non-negotiable rule (read this first)
Every number, percentage, shekel amount, date, or ranking in your verdict MUST come from a source you actually pulled this session (an MCP tool result or a page you fetched) and MUST be quoted with its source name and the data's reference period. If you cannot pull a confirming figure, the verdict is Not enough data (insufficient data). Do not estimate, do not infer from training data, do not "approximately." A fabricated fact-check is the one failure mode this skill exists to prevent.
Step 1: Isolate the checkable claim
Take what the user gave you (often rhetoric plus a fact) and strip it down to the atomic, empirically checkable assertion(s). A single sentence can hold several.
- "This failing government let inflation explode to the highest in a decade" decomposes into: (a) what was the most recent CPI annual rate, (b) is it the highest in ten years. The word "failing" is opinion and gets no verdict.
- Separate fact from opinion from prediction. Opinions ("worst government ever") and predictions ("prices will rise next year") are flagged as out of scope for a verdict. You may show the relevant baseline data, but you do not rate them true or false.
- Causal claims ("the reform caused prices to rise") are not a clean true or false. Data can show a correlation but rarely proves causation. Rate these at your discretion and present the data, do not stamp them true or false.
- A poll claim ("most of the public supports X") is checked against the original poll publication: who ran it, the sample and population, the field dates, the margin of error, and the exact question wording. It is never rated against official data.
- If the claim reached the user inside a screenshot, an image, or a video, the first atomic claim is attribution and authenticity: did this person really post or say it, and is the media unedited and current? Settle that first (Step 3a). If the minister never posted it, the number inside is moot.
Step 2: Recover the original context
The most common reason a true number becomes a misleading verdict is missing context. Before you check, establish: who said it, when, and against which baseline. A figure that is technically correct but compares the wrong years, mixes nominal with real, or quotes a monthly change as if annual, is misleading even when the raw number checks out.
Pin down which figure the claim actually means before you pick a source. Claims are often ambiguous between measures (nominal vs real, this year vs a multi-year total, planned vs executed budget, the Bituach Leumi statutory average wage vs the CBS average monthly wage). Check the most charitable correct reading of the claim, not the measure that is easiest to pull. If the claim is genuinely ambiguous, say so and verify the most plausible reading rather than the one that makes the verdict cleanest.
For "doubled", "tripled", or "jumped by X%" claims, check five things before the figure: percent versus percentage points (a rate that goes from x% to 2x% doubled, but rose by only x percentage points), a low base that makes a small absolute change look dramatic, start and end months chosen to produce the headline, seasonally adjusted versus original series (comparing one of each is a classic misleading), and nominal versus real or per-capita (a multi-year shekel total can double while real spending per person barely moves). CBS publishes both seasonal series for many indicators; compare like with like.
For any "highest, lowest, or first time in N years" claim, you cannot settle it with one current figure plus a remembered peak. Pull the full data series, confirm the index base year did not change mid-series, and compare like for like. A remembered historical peak is a fabricated comparison.
Step 3: Route the claim to the right official source
Match the claim type to its authoritative source. Prefer the dedicated MCP server when installed; otherwise fetch the official page directly. See references/source-map.md for the full table and the per-source pitfalls.
| Claim is about | Authoritative source | MCP (if installed) |
|---|---|---|
| Inflation, CPI, cost of living, a specific price change | Central Bureau of Statistics (CBS) | israel-statistics, israeli-cbs |
| State budget, ministry spending, procurement, support payments | BudgetKey / OpenBudget | budgetkey, il-budget |
| Exchange rate, the value of the shekel | Bank of Israel representative rate | boi-exchange |
| A bill, a committee, what the Knesset passed, how MKs voted | Knesset OData (ParliamentInfo for bills and committees; OData V4 KNS_PlenumVote / KNS_PlenumVoteResult for plenum votes) |
knesset |
| Knesset election results, turnout, party seats | Central Elections Committee (data.gov.il); local-authority elections fall under the Ministry of Interior (it publishes their polling-station data), not the Central Elections Committee | israel-elections |
| Apartment / housing prices, a specific deal | CBS housing index + Nadlan recorded deals | nadlan |
| Rent | CBS CPI rent component (rent), not Nadlan or the House Price Index, which cover purchases | israel-statistics |
| NGO finances, amuta status, support from a foreign state entity | Ministry of Justice Corporations Authority (GuideStar) | israel-amutot |
| Unemployment | CBS labour-force survey | israel-statistics |
| Wages, average wage | CBS average monthly wage + Bituach Leumi statutory average wage | israel-statistics |
| An opinion poll | The original poll publication (institute, sample, dates, margin of error, question wording) | none |
| The Bank of Israel interest rate | Bank of Israel monetary policy announcements | none |
| Poverty | Bituach Leumi annual poverty and income-inequality report | none |
| GDP growth, deficit, national debt | Not yet mapped in this skill: use the official publisher's own release, never a news summary, or return not enough data | none |
| A government failure or a ministry's performance | State Comptroller reports | none |
| Crime, road safety, population, health, vehicles | Israel Police / CBS / Ministry of Health / Ministry of Transport via data.gov.il | data-gov-il, israel-vehicles |
| Anything else with a government dataset | data.gov.il (CKAN search) | data-gov-il, datagov-israel |
Step 3a: Screenshots, images, and video (provenance before data)
When the claim is a screenshot, an image, an audio clip, or a video, the question is provenance, not a data lookup. Work in this order and emit a separate verdict block for "did X post or say this" and for "is the figure true".
- Was it already checked? Search Google Fact Check Explorer (it indexes the Hebrew checks The Whistle publishes) and The Whistle's archive on Globes. For WhatsApp chain letters and rumours, the long-running site irrelevant.org.il (Hanan Cohen) is a useful lead, but it declares itself non-objective, so treat it as a pointer, not an authority. FakeReporter exposes coordinated networks and known fakes. A prior check is a lead: cite it, then re-pull the figure, because data may have been revised since. Absence from any of these proves nothing.
- A screenshot of a post is not evidence the post exists. Look for the original on the person's own verified account, then check the Wayback Machine and archive.today for a capture. If there is neither an original nor a capture, the attribution verdict is not enough data, not false. A missing capture is not proof of fabrication.
- Images and video: run a reverse-image search on the image or on keyframes (Google Lens, TinEye, Yandex Images, Bing Visual Search). The InVID-WeVerify browser plugin splits videos into keyframes for this. Rule out recycled or cut real footage before suspecting synthesis: an old clip, or a clip cut from a longer speech, is missing context, not altered. For a Knesset speech or a ministry statement, find the full original before rating it.
- Suspected AI-generated media: an AI-detector score is not a verdict, and a copy forwarded on WhatsApp is compressed, so reading metadata from it proves little. Ask the user for the earliest source they have. Without the original or a credible provenance trail, the authenticity verdict is not enough data. Forensic attribution (which model, who made it) is out of scope.
Step 4: Pull the figure and record the citation trail
Query the source. For every figure you bring back, record four things, because they go into the verdict verbatim:
- The exact figure as published.
- The source name (e.g. "CBS Consumer Price Index").
- The reference period (e.g. "April 2026", "Q1 2026", "rate of 2026-06-04").
- Any framing caveat (nominal vs real, planned vs executed budget, survey vs registered).
If the source returns nothing, or you cannot reach it, say so plainly and move toward not enough data. Do not substitute a number from memory.
Step 5: Assign one verdict label with the citation trail
Use the scale published by The Whistle (HaMeshurkanit), the Israeli fact-checking desk at Globes and an IFCN signatory since 2018 (its current IFCN verification runs to 23 Jan 2027; re-check after that date). The first five labels below are The Whistle's published scale; the sixth, not enough data, is this skill's anti-fabrication addition for when no source can settle the claim. Use one label per claim:
| Label | When to use it |
|---|---|
| True | The statement is true in its near-entirety. |
| Not accurate | Substantial parts of the statement are wrong. |
| Misleading | The statement creates a false impression or takes facts out of context, even if a raw number is technically correct. |
| False | The statement is false. |
| At your discretion | The factual situation is too complex for a single definitive score. Lay out the data and let the reader judge. |
| Not enough data | No authoritative source confirms or refutes it. This is the anti-fabrication fallback, not a failure. |
For a viral image, audio, or video, Meta's platform labels fit better: false, altered, partly false, missing context, or satire. These are Meta's labels, not The Whistle's: The Whistle left Meta's fact-checking programme in December 2020. In the US, Meta replaced third-party fact-checking with Community Notes; as of September 2026 Meta lists Community Notes only for the US and a test in Latin American countries, so outside those the third-party labels still appear.
Apply the same standard regardless of who made the claim. Non-partisanship is the point.
Step 6: Output the verdict
Produce a compact, quotable verdict block. Default to Hebrew when the claim was Hebrew, English when it was English. Structure:
טענה: <the claim as stated, in one line>
פסיקה: <one label from the scale>
מה הנתונים מראים: <the authoritative figure, quoted>
מקור: <source name>, <dataset / page>, <reference period>, נמשך ב-<date you pulled it>
הערת הקשר: <one line if context changes the picture, otherwise omit>
If the claim decomposed into several atomic assertions (Step 1), emit one verdict block per assertion. A compound claim has no single label: rate each part on its own, and if it helps the reader, add a one-line overall read (for example, partly true and partly misleading). Keep it to the verdict and its evidence. Do not add advocacy, recommendations, or a call to action.
Recommended MCP Servers
Install the MCP that matches the claim type for live, structured access. All are in the agentskills.co.il directory. Without an MCP, fetch the official page directly (see Reference Links), but the MCP is faster and returns structured data you can quote precisely.
| MCP | Use it for |
|---|---|
budgetkey / il-budget |
State budget, executed vs planned spending, procurement, support payments |
israel-statistics / israeli-cbs |
CPI, inflation, housing index, labour force, demographics |
boi-exchange |
Official Bank of Israel representative exchange rates |
knesset |
Bills, committees, member profiles, legislative status |
israel-elections |
Knesset election results, seats, turnout |
nadlan |
Recorded real-estate transaction prices |
israel-amutot |
Non-profit registry, foreign-state-entity support disclosures |
data-gov-il / datagov-israel |
Any government dataset via CKAN search |
Bundled Resources
references/source-map.md: the full claim-type to source mapping, each source's update cadence, and the per-source pitfalls that turn a correct number into a misleading one.references/domain-checklist.md: the coverage contract for this skill, the verdict scale with sources, and the standard fact-check workflow.scripts/verdict_template.py: prints the verdict scale and a fillable verdict skeleton, and suggests the likely source for a claim from keywords, ranked by match strength, flagging screenshots and video as provenance-first. It scaffolds structure only and never produces a figure.
Gotchas
These are mistakes an AI agent makes when fact-checking Israeli claims, not user errors.
- Filling the gap with a plausible number. The single biggest failure. When the source is unreachable or silent, the agent invents a figure that "sounds right". The correct move is not enough data. A fabricated verdict is worse than no verdict.
- Confusing the two unemployment figures. The CBS labour-force survey unemployment rate and the Sherut HaTaasuka count of registered job-seekers are different numbers measuring different things. Quoting one as the other is a classic misleading.
- Planned vs executed budget. A claim about spending must be checked against executed budget (execution), not the original planned budget (original budget). They diverge, and citing the wrong one flips the verdict.
- Market spot rate vs representative rate. For shekel exchange-rate claims, only the Bank of Israel representative rate (representative rate) is authoritative. A market spot rate from a finance site is not the official figure.
- Asking prices vs recorded deals. "Apartment prices" claims need recorded transaction prices (Nadlan) or the CBS House Price Index, not listing/asking prices from a real-estate portal. For a national trend use the CBS index, not an average of individual deals.
- Per-MK plenum votes live in Knesset OData V4, not in ParliamentInfo or the elections committee. The ParliamentInfo OData service exposes bills, committees, and members, but not how each MK voted on the plenum floor. For roll-call votes use the Knesset OData V4 service (
KNS_PlenumVotefor the vote,KNS_PlenumVoteResultfor each MK's vote). EachKNS_PlenumVoteResultrow records one MK's result, and present is not in favour, against, or an abstention; an MK with no row in a vote is not an MK who voted against. The data.gov.il package titled Knesset plenary votes still exists but holds no data files, so it is not a source. Do NOT use the Central Elections Committee data for this: it holds ballot-box election results, not how MKs voted on laws. - Subtracting CPI points across a base change. The CBS rebases its indices, so raw index points from different bases are not comparable. Example: asked for January 2015 to August 2026, the CBS index calculator uses the last published (known) indices for those dates, November 2014 (102.1, on the 2012 base) and June 2026 (104.8, on the 2024 base), and after the chaining coefficient of 1.162 returns a change of 19.3%. Subtracting the raw points would show almost no inflation. Use the calculator or the published chained series, never raw points across a rebase, and state which convention you used: the index for the month itself gives a different result.
- "Foreign-funded NGO" claims. The Ministry of Justice amutot dataset on data.gov.il publishes donations from a foreign state entity (donations from a foreign state entity) as a separate file. That file says nothing about private foreign donors either way, so first establish which kind of funding the claim is about; for the rest use the amuta's filings on GuideStar. Party and candidate funding is a separate regime, under the State Comptroller.
- Provisional vs revised figures. The CBS and the Bank of Israel revise numbers (CPI revisions, GDP revisions, provisional vs final budget execution). A claim that is true against the first provisional print can be false against the revised one. Note whether the figure you pulled is provisional or final, and pin the verdict to that.
- Treating a stale or empty dataset as current. data.gov.il datasets update on wildly different schedules, and some packages have a title but zero data files. Always read the resource's last-updated metadata and file count, and state the reference period; do not present three-year-old data as today's.
Reference Links
| Source | URL | What to check |
|---|---|---|
| The Whistle methodology (verdict scale) | https://www.globes.co.il/news/article.aspx?did=1001372882 | The five Hebrew verdict labels and editorial method |
| IFCN Code of Principles | https://ifcncodeofprinciples.poynter.org/the-commitments | Non-partisanship, source and methodology transparency |
| Central Bureau of Statistics | https://www.cbs.gov.il | CPI, price indices, labour force, housing index, demographics |
| Bank of Israel exchange rates | https://www.boi.org.il/en/economic-roles/financial-markets/exchange-rates/ | Official representative rate (representative rate) |
| BudgetKey / OpenBudget | https://next.obudget.org/ | State budget, executed spending, procurement, support payments |
| Knesset OData | https://knesset.gov.il/Odata/ParliamentInfo.svc/ | Bills, committees, member profiles |
| Knesset OData V4 | https://knesset.gov.il/OdataV4/ParliamentInfo/ | Plenum votes (KNS_PlenumVote, KNS_PlenumVoteResult) |
| Central Elections Committee (data.gov.il) | https://data.gov.il/he/datasets/central-election-committee/votes-knesset | Official election results, seats, turnout |
| Government real-estate site (Nadlan) | https://www.nadlan.gov.il/ | Recorded real-estate transaction prices |
| Bituach Leumi average wage | https://www.btl.gov.il/Mediniyut/GeneralData/Pages/%D7%A9%D7%9B%D7%A8%20%D7%9E%D7%9E%D7%95%D7%A6%D7%A2.aspx | Statutory average wage under the National Insurance Law (wage claims usually mean the CBS average monthly wage instead) |
| FakeReporter | https://fakereporter.net/ | Exposed fakes and coordinated networks |
| Google Fact Check Explorer | https://toolbox.google.com/factcheck/explorer | Whether a claim was already checked |
| Wayback Machine / archive.today | https://web.archive.org/ , https://archive.ph/ | Captures of posts and pages that were deleted or edited |
| TinEye | https://tineye.com/ | Earliest appearance of an image online |
| GuideStar (Ministry of Justice) | https://www.guidestar.org.il/ | Amuta filings and status |
| data.gov.il | https://data.gov.il | CKAN search for any ministry dataset |
Troubleshooting
The source is down or returns nothing. Do not substitute a remembered figure. Retry once, try the dedicated MCP if you used the web page (or vice versa), and if still empty, return not enough data and say which source you could not reach.
The claim mixes several facts. Decompose it (Step 1) and issue a separate verdict per atomic assertion. A compound statement can be partly true and partly false.
The claim is an opinion or a prediction. It gets no true/false verdict. Say so, and optionally present the relevant baseline data without rating it.
Two sources disagree. Prefer the primary official dataset over a news summary of it. If two official sources genuinely differ (e.g. different definitions), that is a for your judgment, and you explain the definitional gap rather than picking one.
The user wants the verdict in the other language. Translate the prose but keep the source names and reference periods intact, and keep the Hebrew verdict label alongside its English gloss.