Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Optimize multilingual product listings for international e-commerce including SEO localization, machine translation workflows, and cultural adaptation. Use this skill when the user needs to create product listings in multiple languages, optimize for local search, or adapt marketing content for different markets — even if they say 'translate our listings', 'optimize for local SEO', 'adapt for the Japanese market', or 'our translated listings don't convert'.
.claude/skills/asgard-ai-platform-ecom-multilingual-listing/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 16% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 33% | 0% |
| case-06 | ✓→✓ | = Same ✓ | 43% | 0% |
| case-07 | ✓→✓ | = Same ✓ | 28% | 0% |
IRON LAW: Localization ≠ Translation
Translating product listings word-for-word produces listings that are
grammatically correct but commercially ineffective. Localization adapts
the message for the TARGET MARKET's search behavior, cultural preferences,
and purchasing psychology.
"機能性飲料" doesn't translate to "functional beverage" for a US audience —
it translates to "energy drink" or "performance drink" because that's
what Americans search for.Phase 1: Keyword Research (per market)
Phase 2: Listing Structure | Element | Localization Requirement | |---------|------------------------| | Title | Include top local keywords, follow platform character limits, front-load important terms | | Bullet points | Highlight benefits that matter to LOCAL consumers (may differ from home market) | | Description | Natural language with local idioms, NOT translated corporate-speak | | Images | Local models/settings, local measurements (cm vs inches), local lifestyle context | | Search terms | Backend keywords in local language, including misspellings and synonyms |
Phase 3: Cultural Adaptation
Phase 4: Quality Assurance
| Tier | Method | Cost | Quality | Use Case | |------|--------|------|---------|----------| | Machine translation | AI (GPT, DeepL) | Lowest | Readable but often unnatural | Internal use, draft | | Machine + human edit | AI draft → native editor | Medium | Good | B-items, high-volume SKUs | | Professional localization | Human translator with market expertise | Highest | Best | A-items, brand-critical content | | Transcreation | Creative rewriting for target culture | Premium | Exceptional | Taglines, brand stories, ads |
markdown# Listing Localization: {Product} → {Target Market} ## Keyword Research | Local Keyword | Search Volume | Competition | Priority | |-------------|-------------|------------|---------| | {keyword} | {N/month} | H/M/L | 1/2/3 | ## Localized Listing - **Title**: {optimized for local keywords} - **Bullet Points**: {adapted for local benefits emphasis} - **Description**: {localized, natural language} ## Cultural Adaptations | Element | Home Market | Target Market | Change | |---------|-----------|-------------|--------| | {element} | {original} | {adapted} | {why} | ## QA Checklist - [ ] Native speaker reviewed - [ ] Keywords included in title/backend - [ ] Sizing chart converted - [ ] Images culturally appropriate
references/marketplace-seo.md| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 108,025 | 25,679 | -76% | 1 | 1 | 0% | 6,401 | 5,186 | -19% | 0 | 0 | — |
case-02 | fail→fail | 26,566 | 14,264 | -46% | 1 | 1 | 0% | 4,739 | 3,481 | -27% | 0 | 0 | — |
case-03 | fail→fail | 41,373 | 30,192 | -27% | 1 | 1 | 0% | 5,459 | 4,955 | -9% | 0 | 0 | — |
case-04 | pass→pass | 17,817 | 13,495 | -24% | 1 | 1 | 0% | 2,421 | 2,804 | +16% | 0 | 0 | — |
case-05 | pass→pass | 20,514 | 22,948 | +12% | 1 | 1 | 0% | 2,897 | 3,859 | +33% | 0 | 0 | — |
case-06 | pass→pass | 16,168 | 17,536 | +8% | 1 | 1 | 0% | 2,267 | 3,231 | +43% | 0 | 0 | — |
case-07 | pass→pass | 35,270 | 12,844 | -64% | 1 | 1 | 0% | 2,244 | 2,880 | +28% | 0 | 0 | — |
case-08 | fail→pass | 16,314 | 18,125 | +11% | 1 | 1 | 0% | 2,033 | 3,333 | +64% | 0 | 0 | — |
case-09 | pass→pass | 11,349 | 10,636 | -6% | 1 | 1 | 0% | 1,467 | 2,550 | +74% | 0 | 0 | — |
case-10 | pass→pass | 17,218 | 15,290 | -11% | 1 | 1 | 0% | 2,341 | 3,254 | +39% | 0 | 0 | — |
case-11 | pass→pass | 13,445 | 10,754 | -20% | 1 | 1 | 0% | 2,048 | 2,643 | +29% | 0 | 0 | — |
case-12 | pass→pass | 13,998 | 13,995 | -0% | 1 | 1 | 0% | 2,072 | 2,742 | +32% | 0 | 0 | — |
case-13 | pass→pass | 15,115 | 12,603 | -17% | 1 | 1 | 0% | 1,808 | 2,888 | +60% | 0 | 0 | — |
case-14 | pass→pass | 18,263 | 16,984 | -7% | 1 | 1 | 0% | 2,763 | 3,414 | +24% | 0 | 0 | — |
case-15 | pass→pass | 15,590 | 14,100 | -10% | 1 | 1 | 0% | 2,020 | 3,173 | +57% | 0 | 0 | — |
case-16 | pass→pass | 17,028 | 10,283 | -40% | 1 | 1 | 0% | 2,243 | 2,558 | +14% | 0 | 0 | — |
case-17 | pass→pass | 12,127 | 12,273 | +1% | 1 | 1 | 0% | 1,696 | 2,830 | +67% | 0 | 0 | — |
case-18 | pass→pass | 21,580 | 18,836 | -13% | 1 | 1 | 0% | 2,498 | 3,350 | +34% | 0 | 0 | — |
case-19 | fail→fail | 16,573 | 18,975 | +14% | 1 | 1 | 0% | 2,503 | 4,467 | +78% | 0 | 0 | — |
case-20 | fail→fail | 10,427 | 11,406 | +9% | 1 | 1 | 0% | 1,373 | 2,411 | +76% | 0 | 0 | — |
case-21 | fail→fail | 15,835 | 29,477 | +86% | 1 | 1 | 0% | 2,701 | 4,989 | +85% | 0 | 0 | — |
case-22 | fail→fail | 18,312 | 21,321 | +16% | 1 | 1 | 0% | 3,260 | 4,176 | +28% | 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. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 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.