Fintech ux
Design skills for AI agents — roast-my-ui, de-slop, fix-contrast & more. Senior design judgment with computed receipts. Install: npx skills add ravipalwe/design-skills
npx -y skills add ravipalwe/design-skills --skill fintech-uxAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Fintech UX — design judgment for money products, encoded from 19+ years of financial services design leadership. Activates whenever the work touches money: banking, payments, checkout and billing, invoicing, insurance, lending, wealth/investing, crypto wallets, payroll, expense tools — building new UI, reviewing existing UI, or designing critical flows (onboarding/KYC, transfers, failed payments, disputes). Covers trust architecture, financial-anxiety-aware design, money display craft (tabular figures, pending vs posted, negative conventions — with a bundled lint that computes violations from code), compliance as a design material (KYC, AML, PCI-DSS, Reg E, TILA), and the anti-patterns that erode trust or invite regulatory risk. Use when the user mentions a financial product or feature, a payment/checkout/billing flow, KYC or verification UX, displaying balances/transactions/rates, or asks "does this feel trustworthy?" — even if they never say the word "fintech".
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
8.4 KB, as published. Nobody here has run it
Fintech UX — Design for Money
When users trust a screen with their money, every pixel carries weight. A misalignment that reads as "a little off" in a social app reads as "I don't trust this" in a banking app. Financial stress changes how people use software: they check balances at 2am, they abandon KYC flows at the first confusing step, they read a failed payment as "where is my money?" — and the interface either answers or loses them.
This skill applies whether you're building money UI, auditing it, or designing a flow. It's judgment, not a checklist: every rule below exists because violating it costs trust, money, or both.
The Seven Laws of Money UI
- Trust is the product. Features are why users arrive; trust is why they stay. Every decision either deposits or withdraws from the trust account — and trust is built in drops, lost in buckets.
- Calm is functional. Financial anxiety is a usability bug, not a mood. Muted palettes, generous space, deliberate motion, and no alarm styling on ordinary events.
- Numbers are sacred. Tabular figures, explicit signs, honest precision, visible "as of" freshness. Never rounding theater, never a jumping digit.
- Say if money moved. Every error, timeout, and ambiguous state answers the user's only real question first: did money leave my account?
- Irreversible actions get explicit confirmation. The CTA names the action and the amount ("Send $250.00 to Alex Rivera"), never "Confirm."
- Compliance is a design material. KYC, disclosures, and consent are moments to design — explained steps convert; walls of forms abandon (Signicat's Battle to Onboard research measured 63–68% abandonment of badly-designed financial onboarding).
- Accessibility is law, not polish. ADA / Section 508 / WCAG 2.2 AA are legal obligations for financial services. 48px targets for money actions, announced amounts, 200% zoom, no color-only status.
Modes — each has a contract, not just a posture
Build — designing or implementing money UI. Load the reference for the surface you're touching (map below) and apply its patterns and tokens. When creating a design system for a money product, the trust/anxiety principles shape the direction, not just components — calm is a token-level decision. Exit criteria — before declaring done: (1) run the relevant reference's audit checklist against your own output and show the pass/fail lines; (2) if you wrote code, run the lint (below) and clear every ERROR. "Done" means both are clean, not "it renders."
Audit — reviewing an existing money product. Run the laws as lenses over the UI, then load references/trust.md and references/anxiety.md for the deep checks. If code was provided, run the lint first — computed findings anchor the review. If roast-my-ui is installed, compose: its ten passes plus these domain lenses make a fintech-grade roast.
Deliverable — always this shape:
## Verdict
<one sentence — the biggest truth about this product's money UX>
## Findings — ranked by severity
<each finding, no exceptions:
**Law/reference violated** → evidence (screen or file:line — quote it)
→ impact (user harm + regulatory exposure, if any) → exact fix (values, copy)>
## Lint receipts
<script output, when code was provided — file:line, computed>
## What's working
<honest strengths to protect while fixing the rest>
## Fix order
<numbered, highest impact-per-effort first, one line each>
Flow — designing a critical money journey (onboarding/KYC, payment, transfer, dispute, recovery). Load references/flows.md and references/compliance.md.
Deliverable — a flow spec, built in this order:
- Unhappy paths first — enumerate every failure, timeout, rejection, and abandonment point before drawing the happy path; each one gets its recovery designed, not just named.
- Per screen: purpose, all states (loading / error / empty included), and the key copy — error copy answers "did money move?" in its first line.
- Every regulated element tagged inline with its regime (
Reg E:,TILA:,KYC:,PCI:) so the next designer doesn't "clean it up." - Close with exit criteria: the compliance.md audit checklist run against the spec, pass/fail shown.
Receipts — compute, don't assert
When the input includes code (CSS/HTML/JSX/components), run the bundled lint before writing a single finding:
python3 scripts/money_display_lint.py <path> [--json]
It mechanically detects: hyphen-minus negatives (want U+2212), symbol–amount gaps, dropped cents, money columns without tabular figures, color-only negative styling, fee/disclosure text below WCAG 4.5:1 (contrast computed, including var() resolution), and hand-rolled formatting that should be Intl.NumberFormat. Cite findings as file:line receipts — "your transaction column uses proportional figures, here are the 4 files" is unanswerable. MD003/MD005 are questions to verify against context, not automatic verdicts. For colors the lint can't reach (screenshots, Figma), sample and verify with roast-my-ui's scripts/color.py when installed.
Reference map — load just-in-time
| Working on | Load |
|---|---|
| Balances, amounts, transactions, rates, account numbers | references/money-display.md |
| Security UX, credibility, trust signals vs theater | references/trust.md |
| Stress moments, loss display, alerts, tone | references/anxiety.md |
| Onboarding/KYC, payments, errors, disputes, waiting | references/flows.md |
| KYC/AML/PCI/Reg E/TILA touchpoints, disclosures, consent | references/compliance.md |
Don't read them all upfront; each is dense and scoped.
Anti-patterns — never, in any mode
These erode trust and several invite regulatory exposure (dark-pattern enforcement is active — CFPB, FTC, EU):
- Hidden fees — costs revealed only at the final step. Fees appear before commitment, always.
- Urgency theater — countdowns and "limited time" pressure on financial decisions.
- Confirm-shaming — "No, I don't want to save money." Never.
- Color-only status — success/pending/failed must carry icon or label, not just hue.
- Alarm styling on ordinary states — a negative balance is information, not a siren. Minus sign + subtle color, no bold red panic.
- Buried safety controls — card freeze, 2FA, fraud reporting reachable in ≤2 taps from anywhere.
- Black-box processing — "being processed" with no timeline, no status, no way to check.
- Trap-door retention — cancel/close flows harder than open flows. Regulators read this as a dark pattern; users read it as a warning.
Composing with the family
- roast-my-ui installed → offer the fintech-grade roast (its crit + these lenses).
- designos installed → money products bias the direction toward calm worlds; token systems gain the money-display tokens;
scripts/color.pythere verifies the contrast this skill demands. - prism installed → for research deliverables (personas, journey maps) on financial products, prism's finserv layer is the research-side counterpart to this design-side skill.
None are required — this skill stands alone.