Using the compliance skills.

How to run an operator-grade check inside your own AI tool — and how to read what comes back.

The one-sentence version.

Give the model your marketing asset and ask for the check by name — “Run the MiCA marketing self-audit on this banner copy” — and the skill does the rest: every rule applied, every finding quoted, every fix suggested.

What to check.

Anything that reaches a customer: ads, landing pages, emails, X posts, KOL briefs and contracts, video scripts, app-store listings, press releases. One asset at a time gives the sharpest findings — paste the text, or attach the file or screenshot where your tool reads images.

Multi-market asset? Run each target regime in turn — an EU + UK banner should pass both the MiCA and FCA packs before it ships anywhere.

Reading the verdict.

Every check returns a verdict, per-rule findings with the triggered line quoted back, and rewrite suggestions. The three levels mean exactly this:

FAIL

The enforcement class. This marketing should not ship as-is — fix the quoted line, then re-run the check on the fixed copy.

FLAG

What an examiner reads first. Not automatically unlawful — but worth fixing before shipping, and worth a note explaining the judgment if you keep it.

PASS

The obvious problems are absent. A prompt to stop worrying — not a certification, and never a substitute for judgment on what the rules can’t see.

The workflow that sticks.

Make the check a pre-flight gate, not an afterthought: draft → check → fix → re-check → ship. Keep the finding and its date with the asset — that record is your audit trail when a regulator, a partner or a board asks how marketing gets reviewed. Re-run whenever the asset changes or the pack’s last-reviewed date moves.

Prompts that work.

Run the MiCA self-audit on this landing-page copy: [paste]

Check this thread against the FCA financial-promotions pack.
UK retail audience, cryptoasset product.

This banner ships in Germany and France next week — full MiCA
check. List every flag with the rule quoted, then rewrite the
copy so it passes.

The more context you give — audience, market, what the product actually is — the sharper the judgment calls on the borderline rules.

Common questions.

Which pack for which market?
EU → MiCA (add GDPR for anything touching email or personal data). UK → FCA. US → SEC. Singapore → MAS. Dubai → VARA. Token launch → TGE readiness. Paid banners and creatives → ad creative. AI-search visibility → AEO. Multi-market assets run each regime in turn.
How is this different from the free checks on this site?
The free checks run a 5-rule sample in the browser — a fast first read. The packs are the full rule sets (267 rules across nine packs), run inside your own tools, on unlimited assets, with your own context and files in play.
Which model should I run the packs on?
The strongest one you have — the packs are model-agnostic Markdown, but rule judgment scales with model quality. A frontier model applying 40 rules to a nuanced landing page is a materially better reviewer than a small one.
Can the check be wrong?
Yes — it’s a review, not an oracle. Treat FAIL findings as near-certain problems and PASS as the absence of obvious ones. The packs state what each rule says and quote your triggering line, so every finding can be verified against the regulation itself — and structural questions still belong with counsel.

Know before a regulator does.

Paste a URL or your copy — the full verdict lands in seconds. Free, no signup.

Run a free check →