Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Fact Checker: Verify Every Claim Before You Publish

    1

    Fact Checker makes your agent work like a newsroom verification desk.

    Free

    7 uses5.0 (1 review)
    Fact Checker: Verify Every Claim Before You Publish

    Fact Checker: Verify Every Claim Before You Publish

    Example session with this skill installed

    Fact-check this paragraph from our blog before we publish: "Spain is the largest e-commerce market in Europe, with online sales growing 40% last year. According to a 2019 study, 9 out of 10 customers abandon a brand after one late delivery."

    • Read your context and instructions
    • Compiled the fact checker

    Example output, abbreviated. Sources shown as placeholders.

    Summary

    3 claims checked (full mode): ❌ 1 incorrect · 🌀 1 misleading · ❓ 1 unverifiable.
    Biggest problems: (1) "largest market in Europe" is wrong, (2) the 40 % figure mixes a sub-category with the whole market, (3) the "9 out of 10" study can't be traced to an original source.

    Claims

    #ClaimVerdictWhat's actually trueSource
    1"Spain is the largest e-commerce market in Europe"❌ IncorrectThe UK, Germany and France have larger online sales; Spain ranks below them[1] European e-commerce association report, latest edition (A/B)
    2"online sales growing 40% last year"🌀 MisleadingThe 40 % refers to one product category in the source; total online sales grew much less[2] national statistics office / competition authority e-commerce data, latest quarter (A)
    3"a 2019 study: 9 out of 10 customers abandon a brand after one late delivery"❓ UnverifiableWidely repeated in blogs; every citation chain ends at a vendor press release without methodology. Searched: exact phrase, 2019 delivery surveys, vendor sites[3] vendor press release (C) - not the original study

    Corrected text

    Spain is the largest e-commerce market in Europe

    Spain is one of Europe's major e-commerce markets [1], with online sales growing 40% last year

    online sales growing [x] % last year [2]. According to a 2019 study, 9 out of 10 customers abandon a brand after one late delivery. Surveys suggest late deliveries are a major reason customers switch brands [4].

    Watch list

    • Growth figures change every quarter - update before publishing next year
    • Claim 3 appears in many competitor blogs; citing it would hurt credibility

    Verified with 6 searches (web search with sources, 6 calls, ~$0.02). Replace [x] with the exact figure from source [2] - confirm which period your post refers to.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Publish with confidence, not "probably fine"

    AI-written drafts confidently invent statistics, misquote people and cite sources that don't say what they claim. Even human writing repeats outdated numbers. Fact Checker makes your agent work like a newsroom verification desk.

    What it does

    • Extracts every checkable claim with the exact quote: statistics, dates, names and titles, quotes, cause-effect claims, superlatives ("first", "largest"), definitions, legal/medical/financial statements, product claims. Flags opinions presented as facts.
    • Two modes: quick (key numbers, names, dates, quotes) or full (every factual statement).
    • Goes to the original: the study, dataset, filing, law, official page or the speaker's own words, not articles repeating them. Checks units, periods, scope, context and newer data.
    • Evidence bar: one authoritative primary source or two independent reputable ones for key claims; never relies on memory for facts that change.
    • Six clear verdicts: correct, imprecise, outdated, misleading, incorrect, unverifiable (with what was searched).
    • Report: summary with counts and the 3 biggest problems, a claims table with sources and dates, a corrected version with minimal tracked edits in the author's voice, and a watch list of claims likely to go stale.
    • Built-in methods by claim type, red flags (citations that don't contain the claim, unit or scope drift, unsourced round numbers), search patterns for primary documents.
    • Web search on Agensi: can use a cheap search-with-sources tool (fractions of a cent per call) when the agent has no web access; the agent shows the price and asks first.
    • Neutral and honest: evidence over opinion, never invents sources, recommends expert review for medical, legal and financial claims.

    Includes

    • SKILL.md workflow
    • references/methods.md - source hierarchy, checks per claim type, red flags, search patterns, correction style

    Pairs with Content Writer and Deep Research. Works with Claude Code, Claude (Cowork), Codex, Cursor and any SKILL.md-compatible agent.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Research & Analysis.

    Opinions and predictions are skipped. Requires access to primary sources via web search. Claims with no public documentation will be marked as unverifiable.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    5.0
    1 review
    5
    1
    4
    0
    3
    0
    2
    0
    1
    0

    This skill has been used 7 times.

    Trust & safety

    Security scan

    Security scan passed 2 days ago

    • Free with an Agensi account

    Permissions

    Network Access
    Listed2 days ago
    Updatedtoday

    What's inside

    Frequently Asked Questions

    Unlike general AI prompts, this skill enforces a 'primary source first' protocol, cross-referencing claims against government databases and official filings to eliminate the 'echo chamber' effect of AI hallucinations.

    The skill identifies and labels every checkable claim—including statistics, dates, and technical figures—with a specific verdict such as 'Correct,' 'Outdated,' or 'Misleading.'

    This skill is designed for cross-platform compatibility and can be integrated into any AI agent or workflow that supports web-enabled research tools and text processing.

    No, this skill focuses strictly on verifiable data; subjective opinions, stylistic choices, and future predictions are excluded from the audit to ensure the report remains objective.

    If a claim involves private data or niche information not available in public records, the skill will transparently mark it as 'Unverifiable' rather than guessing.

    You receive the complete skill configuration including the claim extraction logic, the primary source verification framework, and the structured reporting template for immediate deployment.

    New in Research & Analysis

    Free