Use when starting a launch review, and again whenever features change, to detect whether a product touches a regulated or high-risk domain: health, finance and payments, education, children, legal or professional advice, employment, insurance, crypto, gambling, alcohol or cannabis, biometrics, precise location, or automated decisions. It stops ordinary web-compliance assumptions and routes to human specialist review. Do not use it to interpret those regimes or to clear a product as unregulated.
Installs into .claude/skills of the current project.
Are you the author of Regulated Domain Triggers?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/moh-obaida-regulated-domain-triggers)
---
name: regulated-domain-triggers
description: "Use when starting a launch review, and again whenever features change, to detect whether a product touches a regulated or high-risk domain: health, finance and payments, education, children, legal or professional advice, employment, insurance, crypto, gambling, alcohol or cannabis, biometrics, precise location, or automated decisions. It stops ordinary web-compliance assumptions and routes to human specialist review. Do not use it to interpret those regimes or to clear a product as unregulated."
license: Apache-2.0
metadata:
kind: specialist
compliance-domains: "11"
launch-checks: "6"
references: "official-sources"
---
# regulated-domain-triggers
General web compliance (a privacy page, a consent banner, a sitemap) does not cover a health-data product, a lender, or a children's game. This skill is a **tripwire**: it notices when the product has moved out of ordinary territory and says so *before* anyone reassures the owner with a checklist.
## Activate when
- Recon runs on a new project, or product scope changes (new feature, new data type, new audience).
- Product copy, fields, integrations, or categories suggest any domain below.
- Not to explain those regimes or to say they don't apply.
## Inspect
Search copy, routes, schemas, form fields, dependencies, integrations, and imagery for signals. **One signal is enough to escalate**; absence of signals is *not* proof of absence.
| Domain | Signals |
|---|---|
| Health / wellness | symptoms, diagnoses, medication, therapy, fitness data, wearable sync, patient/clinic/appointment, HIPAA-like words, mental-health, fertility, allergy |
| Finance / payments | lending, credit, investing, trading, brokerage, banking, KYC/AML, bank-account linking (Plaid), payouts/marketplace money movement, tax, insurance quotes, invoicing with custody of funds |
| Children / education | ages under 18, school, students, teachers, grades, classroom, parent portal, kids' content or games, edtech LMS |
| Legal / professional advice | legal, medical, tax, or financial advice, contracts as a service, "attorney", "licensed" claims |
| Employment / HR | hiring, screening, résumés, performance, background checks, gig workers |
| Crypto / gambling / restricted goods | tokens, wallets, exchanges, staking; betting, casino, lottery; alcohol, tobacco/vape, cannabis, weapons, adult content |
| Sensitive data | biometrics, face/voice, precise location, government IDs, race/ethnicity, religion, sexual orientation, union status, criminal records |
| Automated decisions / AI | scoring, ranking people, eligibility decisions, AI advice, AI chat with users (`ai-features-readiness`), synthetic media |
| Public-sector / critical | government customers, procurement forms, safety-critical, elections |
Record: domain, signal (file/route/copy), and what the product does with it.
## Evidence that counts
Label each claim OBSERVED, SOURCE-INDICATED, DECLARED, INFERRED, UNKNOWN, or REVIEW REQUIRED. UNKNOWN is never a pass and never a failure.
- A signal in copy is OBSERVED text. Whether the *regime* applies is REVIEW REQUIRED, always.
- "Wellness journal" with symptom tracking is health-adjacent; do not decide where the line is.
- **Legal specifics: never from memory.** When a rule, deadline, threshold, or required wording matters, read the current text or guidance at an official source while you run (start from [references/official-sources.md](references/official-sources.md)), cite the source and access date, and treat applicability to this business as REVIEW REQUIRED. If you cannot look it up, the answer is UNKNOWN.
## May change
Only `.readyvibe/context.md` ("Regulated-domain flags"). Do not change product code. Do not soften copy to dodge a category ("we're not a bank") since that is a legal and honesty question for the owner.
## Must not claim
"Not regulated", "outside HIPAA/GLBA/FERPA/COPPA/PSD2/MiCA/…", "no license needed", or that ordinary privacy/terms/cookie work is sufficient here. Never advise on licensing or regime scope.
## Verify
Search a second way before concluding: read the routes and schema, the visible copy and metadata, the dependencies and integrations, and the forms; a flag or a "none found" should survive all four. Each flag must cite its signal (file, route, or copy) so the owner can check it, and the report must say which places were searched. Re-run when features change. Ensure the relevant lanes are told: legal-page drafting (`privacy-policy`, `terms-of-service`), analytics/replay (`analytics-privacy`), and `consumer-protection-readiness` must treat this surface as REVIEW REQUIRED.
## Escalate
Every flag. Tell the owner plainly: "This product appears to touch <domain>. Ordinary launch checks do not cover the rules that may apply. Obtain qualified legal/compliance review before launch. I have not assessed those rules." Stop ordinary compliance work on the flagged surface (do not generate a generic policy for it) and continue technical work that is safe regardless (broken links, metadata, headers).
## No change is valid when
No signals were found in the searched places: record "no regulated-domain signals found in <places>; recheck when features, data types, or audience change." This is not clearance.
