Install any skill in seconds. Free to start, no credit card required.
Get Started Free →When the user wants to create or optimize a refund or return policy page. Also use when the user mentions "refund policy," "return policy," "money-back guarantee," "returns and refunds," "refund page," "return process," "refund terms," "satisfaction guarantee," "cancellation policy," or "withdrawal right." For legal overview, use legal-page-generator.
.claude/skills/kostja94-refund-page-generator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 85% | 0% |
Guides refund and return policy page content for SaaS, digital products, e-commerce, and services. Covers jurisdiction-specific requirements including EU withdrawal rights.
When invoking: On first use, if helpful, open with 1–2 sentences on what this skill covers and why it matters, then provide the main output. On subsequent use or when the user asks to skip, go directly to the main output.
Identify:
Refund policies differ radically by product type. Identify the category first:
| Category | Default Refund Stance | Key Considerations | |---|---|---| | SaaS / Subscription | Pro-rata or conditional | EU withdrawal button required by Jun 2026; trial periods; auto-renewal cancellation; Sky Austria case may reclassify subscriptions as "digital services" with full-period withdrawal rights | | Digital Downloads | Typically non-refundable | "Once downloaded, all sales final"; exceptions for technical failure or duplicate purchase; must obtain explicit consent to waive EU withdrawal right before download | | Physical Goods | 30-day return window (standard) | Return shipping costs; restocking fees; original packaging; final-sale categories | | Courses / Coaching | Conditional | Pre-recorded: non-refundable once accessed; Live: partial refund before start, none after; rescheduling preferred over refunds | | API / Usage-Based | Prorated or non-refundable | Pay-as-you-go typically non-refundable; subscription components may be prorated | | Marketplace | Per-vendor or uniform | Who processes refunds; vendor return rates (>8% review, >15% suspend); uniform policy builds trust | | Free Tools | N/A | No payment = no refund policy needed; state "free, no purchase required" if asked |
14-Day Cooling-Off Period:
Mandatory Withdrawal Button — Deadline June 19, 2026 (Directive EU 2023/2673):
This is the most urgent 2026 compliance item. Requirements:
Digital Content vs Digital Services — Sky Austria Case (CJEU, pending 2026):
This pending case may significantly change how SaaS refunds work:
Exceptions (Article 16 CRD — withdrawal right does NOT apply to):
2-Year Conformity Guarantee: Under Digital Content Directive (2019/770), consumers have a minimum 2-year guarantee for defective digital products and services.
Default: "Due to the digital nature of this product, all sales are final and non-refundable."
Recommended Exceptions:
EU Compliance for Digital Products:
Content License Clause: Make clear the buyer owns a license to use, not the content itself. "You are purchasing a non-exclusive, non-transferable license to use product], not ownership of the underlying content."
Every refund/return policy should cover:
| Element | What to Include | |---------|----------------| | Eligibility | What can be returned/refunded; time limits; condition requirements | | Non-refundable items | Final-sale categories; digital products with access/waiver | | Refund amount | Full, pro-rata, minus fees, store credit only | | Refund method | Original payment, store credit, check | | Refund timeline | When the customer will receive the refund | | Process | Step-by-step instructions; withdrawal button location; contact | | Exceptions | Defective items, duplicate purchases, technical failures | | Trial terms (SaaS) | Trial duration, conversion date, billing after trial | | Cancellation (SaaS) | How to cancel; what happens to data; pro-rata refund details | | Jurisdiction rights | EU 14-day withdrawal; Australia consumer guarantees; other statutory rights |
Place the refund policy where customers need it before making a purchase decision:
For SaaS: the cancellation flow should be accessible from account settings, billing page, and support contact — never require calling support as the only cancellation method.
| Metric | Benchmark | |--------|-----------| | Median physical-goods return window | 30 days from delivery | | Brands offering free returns (stated on policy page) | 18.4% | | Brands with restocking fees | 19.2% | | Brands requiring original packaging | 50.1% | | Brands with final-sale clauses | 39.9% | | Brands using store-credit-only refunds | 3.3% | | SaaS pro-rata refund standard | Unused months minus processing fee or no refund after window |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 33,314 | 16,559 | -50% | 1 | 1 | 0% | 5,660 | 6,001 | +6% | 0 | 0 | — |
case-02 | fail→fail | 21,676 | 16,773 | -23% | 1 | 1 | 0% | 3,886 | 5,948 | +53% | 0 | 0 | — |
case-03 | fail→fail | 30,295 | 20,619 | -32% | 1 | 1 | 0% | 5,135 | 6,339 | +23% | 0 | 0 | — |
case-04 | pass→pass | 15,251 | 11,298 | -26% | 1 | 1 | 0% | 2,716 | 5,072 | +87% | 0 | 0 | — |
case-05 | fail→pass | 18,807 | 9,773 | -48% | 1 | 1 | 0% | 3,056 | 4,768 | +56% | 0 | 0 | — |
case-06 | pass→pass | 10,827 | 11,579 | +7% | 1 | 1 | 0% | 1,782 | 5,009 | +181% | 0 | 0 | — |
case-07 | fail→pass | 47,556 | 11,584 | -76% | 1 | 1 | 0% | 2,732 | 4,782 | +75% | 0 | 0 | — |
case-08 | pass→pass | 11,226 | 9,142 | -19% | 1 | 1 | 0% | 1,831 | 4,489 | +145% | 0 | 0 | — |
case-09 | pass→pass | 10,583 | 8,235 | -22% | 1 | 1 | 0% | 1,833 | 4,389 | +139% | 0 | 0 | — |
case-10 | fail→pass | 16,312 | 15,712 | -4% | 1 | 1 | 0% | 2,703 | 4,992 | +85% | 0 | 0 | — |
case-11 | pass→pass | 12,894 | 10,403 | -19% | 1 | 1 | 0% | 1,968 | 4,790 | +143% | 0 | 0 | — |
case-12 | pass→pass | 12,307 | 10,631 | -14% | 1 | 1 | 0% | 2,094 | 4,823 | +130% | 0 | 0 | — |
case-13 | pass→pass | 10,883 | 6,247 | -43% | 1 | 1 | 0% | 1,822 | 4,017 | +120% | 0 | 0 | — |
case-14 | fail→pass | 15,501 | 11,147 | -28% | 1 | 1 | 0% | 2,604 | 4,820 | +85% | 0 | 0 | — |
case-15 | pass→pass | 8,977 | 7,320 | -18% | 1 | 1 | 0% | 1,649 | 4,279 | +159% | 0 | 0 | — |
case-16 | fail→pass | 12,436 | 11,224 | -10% | 1 | 1 | 0% | 2,308 | 5,033 | +118% | 0 | 0 | — |
case-17 | fail→fail | 16,537 | 11,063 | -33% | 1 | 1 | 0% | 2,637 | 4,863 | +84% | 0 | 0 | — |
case-18 | pass→pass | 13,131 | 9,763 | -26% | 1 | 1 | 0% | 2,094 | 4,734 | +126% | 0 | 0 | — |
case-19 | pass→pass | 13,364 | 9,361 | -30% | 1 | 1 | 0% | 2,123 | 4,457 | +110% | 0 | 0 | — |
case-20 | pass→fail | 9,221 | 9,272 | +1% | 1 | 1 | 0% | 1,535 | 4,488 | +192% | 0 | 0 | — |
case-21 | fail→pass | 15,736 | 12,772 | -19% | 1 | 1 | 0% | 2,321 | 5,133 | +121% | 0 | 0 | — |
case-22 | pass→pass | 15,399 | 10,161 | -34% | 1 | 1 | 0% | 2,401 | 4,491 | +87% | 0 | 0 | — |
case-23 | pass→pass | 20,592 | 19,529 | -5% | 1 | 1 | 0% | 3,749 | 6,279 | +67% | 0 | 0 | — |
case-24 | pass→pass | 18,017 | 15,447 | -14% | 1 | 1 | 0% | 2,846 | 5,284 | +86% | 0 | 0 | — |
case-25 | pass→pass | 17,883 | 19,629 | +10% | 1 | 1 | 0% | 3,281 | 6,693 | +104% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted. The headline lift of +24 percentage points is the difference between those two pass rates over the 25 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.