Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Pre-Publish Audit: Catch Risky Claims Before You Post

    1

    Paste a post, email, ad or page. A second AI checks it for unproven claims, fake urgency, missing disclosures and lines that invite an easy joke, then returns a verdict with the exact fix. Free.

    Free

    Security scanned
    Pre-Publish Audit: Catch Risky Claims Before You Post

    Pre-Publish Audit: Catch Risky Claims Before You Post

    Example session with this skill installed

    Audit this before I post

    Subject: LAST CHANCE: our best-selling cold brew is almost gone ☕

    Our new Midnight Cold Brew is the smoothest coffee you'll ever taste. 2x the caffeine of regular coffee and zero bitterness. Loved by 10,000+ customers.

    Only 50 bags left, so order before Friday!

    Coming soon: oat-milk concentrate.

    Unlike those burnt chain-store roasts, ours is small-batch.

    • Read your context and instructions
    • Compiled the pre-publish audit

    Verdict: BLOCKED. 3 blocks, 4 fixes, 1 note. No claim ledger was shared, so every product claim is held until you give a source.

    Audit · Midnight Cold Brew email · sample

    Auditor: launch-gatekeeper (independent of: the writer)
    Verdict: BLOCKED

    Claims inventory

    #Claim as writtenLedger statusFinding
    1"best-selling"No sourceBLOCK 2
    2"smoothest coffee you'll ever taste"UnprovableFIX 4
    3"2x the caffeine of regular coffee"No sourceBLOCK 1
    4"zero bitterness"MeasurementFIX 1
    5"Loved by 10,000+ customers"No sourceBLOCK 2
    6"Only 50 bags left... before Friday"UnconfirmedBLOCK 3
    7"Coming soon: oat-milk concentrate"No stageFIX 3

    Findings

    ✗

    BLOCK 1 · body, line 1 · "2x the caffeine of regular coffee" is an unvalidated claim, and "regular coffee" isn't defined. Fix: remove it, or give the source (lab test or roast spec) and name the comparison, e.g. "about twice the caffeine of our drip coffee per 8 oz".

    ✗

    BLOCK 2 · subject + body · "best-selling" and "Loved by 10,000+ customers" have no source. Block both together: fixing one and keeping the other still implies it. Fix: remove both, or send order data and state what it shows.

    ✗

    BLOCK 3 · subject + body · "almost gone" and "Only 50 bags left" are urgency claims. Fix: keep them only if the count is true on send day; otherwise use "This batch ships Friday."

    ⚠ FIX 1 · body · "zero bitterness" is a measurement posing as description. Fix: "a smooth, chocolatey finish".

    ⚠

    FIX 2 · body · "Unlike those burnt chain-store roasts" is an implied competitor dig. Fix: cut the comparison and keep "Ours is small-batch."

    ⚠

    FIX 3 · body · "Coming soon" with no stage or date. Fix: "We're testing an oat-milk concentrate. Want a sample when it's ready? Reply YES."

    ⚠

    FIX 4 · body · "smoothest coffee you'll ever taste" fails the comment-section test (it invites "challenge accepted" replies). Fix: "Our smoothest cold brew yet."

    · NOTE 1 · subject · the ☕ emoji is your call; check how it renders in your email tool.

    Sources needed before this can send

    • Caffeine comparison (lab test or roast spec)
    • Order data behind "best-selling" and "10,000+ customers"
    • Stock count on send day

    Send those and say "re-run the audit". Without them, remove claims 1, 3 and 5 and the email can pass with fixes.

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

    About this skill

    The comment section will find what's wrong with your post. This finds it first. Paste your copy and get a verdict, PASS, PASS WITH FIXES or BLOCKED, with the exact fix for every problem.

    Example

    Your draft: "The best candle you'll ever buy. Burns 50+ hours. Only 3 left!"

    Verdict: BLOCKED

    "The best candle you'll ever buy": a superlative you can't prove. Fix: say what makes it good, like "hand-poured soy, cotton wick."

    "Burns 50+ hours": no test on file. Fix: send your burn-test notes, or use "long-burning."

    "Only 3 left!": is that true right now? If you don't track stock, cut it.

    What it checks

    15 points, in order of what bites most: unproven claims, measurements posing as description, banned phrases, the hidden "coming soon", implied competitor digs, the comment-section test, tagline rules, promises you can't keep, pricing and urgency honesty, testimonial disclosures, regulated claims (health, finance, environmental, children's products), platform ad rules, voice, accessibility, and consistency across a campaign.

    Why it works

    Self-review checks a draft against what you meant. This checks it against what it actually says. Run it in a different chat from the one that wrote the copy, so the read is fresh.

    It's the gatekeeper from the One-Person Marketing Department, where it checks every draft a 30-role AI team writes. Built by Tom Sun, who grew a consumer brand's revenue from about $3M to $6M as its senior director of marketing, and co-founded PAVE Agency (120+ brands).

    How to use it

    Say "Audit this before I post: [paste]". If you keep a list of approved claims or brand rules, share it and the audit checks against it. If you don't, it holds every product claim until you give a source.

    Works with Claude.ai, the Claude desktop app, Cowork and Claude Code. It isn't legal advice: regulated claims are flagged for you to check with the right person.

    It can't check facts on its own: claims are checked against the approved-claims list or sources you give it, and anything unsourced is held, not approved. Regulated claims (health, finance, environmental, children's products) are flagged for you to check; it isn't legal advice. Works best when a different chat or agent runs the audit than the one that wrote the copy.

    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 any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 18 days ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed18 days ago
    Updated5 days ago

    What's inside

    Frequently Asked Questions

    Proofreading checks spelling and flow. This runs a fixed 15-point checklist for the problems that cause trouble after you post: unproven claims, fake urgency, missing disclosures, implied competitor digs and lines that invite an easy joke. Every finding comes with exact replacement wording. Run it in a different chat from the one that wrote the copy, so it reads what the copy says, not what you meant.

    No. Paste the copy and it audits against the checklist, holding every product claim until you give a source. If you keep a list of approved claims and brand rules, share it and the audit checks against that too.

    No. It flags regulated claims (health, finance, environmental, children's products) and missing endorsement disclosures so you can check them with the right person. It helps you catch problems early; it doesn't replace a lawyer.

    A verdict (PASS, PASS WITH FIXES or BLOCKED), a table of every checkable claim in the draft, then each finding marked BLOCK (can't go out as written), FIX (should change before approval) or NOTE (your call), each with the exact fix.

    One skill: the pre-publish-audit SKILL.md, the full 15-point checklist, the launch-gatekeeper role file, a README and the license. No scripts and no setup.

    No. It can't confirm on its own that a claim is true. It checks the copy against the approved claims and sources you give it, and holds anything without a source until you supply one.

    New in Marketing & SEO