Skill Wiki
I18n Launch V1
Accumulated execution feedback for first-release i18n rollouts using @community/collection-i18n-ready-component.…
$ prime install @community/feedback-i18n-launch-v1Projection
Always in _index.xml · the agent never has to ask for this.
I18nLaunchV1 [feedback] v1.0.0
Accumulated execution feedback for first-release i18n rollouts using @community/collection-i18n-ready-component. Synthesized from observed launches at multiple SaaS companies (anonymized) — captures the failure modes that consistently emerge between week 0 and month 6 of multi-locale support.
Loaded when retrieval picks the atom as adjacent / supporting.
I18nLaunchV1 [feedback] v1.0.0
Accumulated execution feedback for first-release i18n rollouts using @community/collection-i18n-ready-component. Synthesized from observed launches at multiple SaaS companies (anonymized) — captures the failure modes that consistently emerge between week 0 and month 6 of multi-locale support.
Target
@community/collection-i18n-ready-component
Observations
- Sample Size: 12
- Period: 2024-Q3 to 2026-Q1
- Success Rate: 0.67
Common Failures:
- Step: @community/principle-no-string-concat
- Failure: early-stage strings ship with concatenation; developer says 'we'll fix it for translation'; never gets fixed and breaks 5 locales
- Frequency: 0.83
- Impact: translation pipeline rejects strings; engineering re-work to refactor; ship date slips 2-4 weeks
- Mitigation: lint rule prohibits string concatenation in JSX/template files from day 1, even before second locale exists
- Step: @community/fact-bidi-direction
- Failure: RTL added late as 'a CSS toggle' — chevrons don't mirror, slide-in animations come from wrong direction, scrollbars on wrong side
- Frequency: 0.75
- Impact: Arabic users see broken UI; bug bash adds 2-3 weeks of layout work post-launch
- Mitigation: audit every directional icon and animation BEFORE first RTL locale; use CSS logical properties (margin-inline-start) instead of physical from start
- Step: translator hand-off
- Failure: translators receive strings without context (no description field, no screenshots) — translations are grammatically correct but contextually wrong
- Frequency: 0.58
- Impact: user complaints about awkward phrasing; multiple revision cycles per locale
- Mitigation: every extracted string includes
descriptionfield + Lokalise/Phrase screenshots auto-uploaded from CI
- Step: pseudo-locale testing
- Failure: team did not run pseudo-locale tests; truncation issues found by real translators in production
- Frequency: 0.5
- Impact: buttons clip text in German (often 30% longer than English); dashboard cards overflow
- Mitigation: add pseudo-locale to CI; visual regression catches truncation before real translation begins
- Step: @community/transform-relative-time-format
- Failure: team rolled own date formatting because Intl.RelativeTimeFormat 'wasn't available everywhere' — actually available in Node 12+ and all modern browsers
- Frequency: 0.42
- Impact: wrong plurals in Russian, Arabic, Polish; embarrassing bugs noticed by community
- Mitigation: use Intl.* APIs unconditionally; check support matrix once and trust it
- Confidence Adjustments:
- @Community/Principle No String Concat: 0.95
- @Community/Pattern Icu Message Format: 0.92
- @Community/Fact Bidi Direction: 0.78
- @Community/Rule Cjk Line Break: 0.72
- @Community/Anti Pattern Flag As Language: 0.96
Ai Confidence
0.81
Lessons
- Day-one i18n hygiene (no concatenation, ICU templates, dir-aware CSS) costs near zero. Retrofitting at month 6 costs quarters.
- Translators are users — give them context, screenshots, and tooling. Strings without context produce wrong translations.
- Pseudo-localization in CI catches 80% of truncation/overflow issues before any real translator sees the build.
- RTL is a layout problem more than a string problem. Audit it before adding the first RTL locale, not after.
- Multi-locale launch should be staged: first 2 locales (often en + es or en + fr) prove the pipeline. Only then expand.
Loaded when retrieval picks the atom as a focal / direct hit.
I18nLaunchV1 [feedback] v1.0.0
Accumulated execution feedback for first-release i18n rollouts using @community/collection-i18n-ready-component. Synthesized from observed launches at multiple SaaS companies (anonymized) — captures the failure modes that consistently emerge between week 0 and month 6 of multi-locale support.
Target
@community/collection-i18n-ready-component
Observations
- Sample Size: 12
- Period: 2024-Q3 to 2026-Q1
- Success Rate: 0.67
Common Failures:
- Step: @community/principle-no-string-concat
- Failure: early-stage strings ship with concatenation; developer says 'we'll fix it for translation'; never gets fixed and breaks 5 locales
- Frequency: 0.83
- Impact: translation pipeline rejects strings; engineering re-work to refactor; ship date slips 2-4 weeks
- Mitigation: lint rule prohibits string concatenation in JSX/template files from day 1, even before second locale exists
- Step: @community/fact-bidi-direction
- Failure: RTL added late as 'a CSS toggle' — chevrons don't mirror, slide-in animations come from wrong direction, scrollbars on wrong side
- Frequency: 0.75
- Impact: Arabic users see broken UI; bug bash adds 2-3 weeks of layout work post-launch
- Mitigation: audit every directional icon and animation BEFORE first RTL locale; use CSS logical properties (margin-inline-start) instead of physical from start
- Step: translator hand-off
- Failure: translators receive strings without context (no description field, no screenshots) — translations are grammatically correct but contextually wrong
- Frequency: 0.58
- Impact: user complaints about awkward phrasing; multiple revision cycles per locale
- Mitigation: every extracted string includes
descriptionfield + Lokalise/Phrase screenshots auto-uploaded from CI
- Step: pseudo-locale testing
- Failure: team did not run pseudo-locale tests; truncation issues found by real translators in production
- Frequency: 0.5
- Impact: buttons clip text in German (often 30% longer than English); dashboard cards overflow
- Mitigation: add pseudo-locale to CI; visual regression catches truncation before real translation begins
- Step: @community/transform-relative-time-format
- Failure: team rolled own date formatting because Intl.RelativeTimeFormat 'wasn't available everywhere' — actually available in Node 12+ and all modern browsers
- Frequency: 0.42
- Impact: wrong plurals in Russian, Arabic, Polish; embarrassing bugs noticed by community
- Mitigation: use Intl.* APIs unconditionally; check support matrix once and trust it
- Confidence Adjustments:
- @Community/Principle No String Concat: 0.95
- @Community/Pattern Icu Message Format: 0.92
- @Community/Fact Bidi Direction: 0.78
- @Community/Rule Cjk Line Break: 0.72
- @Community/Anti Pattern Flag As Language: 0.96
Ai Confidence
0.81
Lessons
- Day-one i18n hygiene (no concatenation, ICU templates, dir-aware CSS) costs near zero. Retrofitting at month 6 costs quarters.
- Translators are users — give them context, screenshots, and tooling. Strings without context produce wrong translations.
- Pseudo-localization in CI catches 80% of truncation/overflow issues before any real translator sees the build.
- RTL is a layout problem more than a string problem. Audit it before adding the first RTL locale, not after.
- Multi-locale launch should be staged: first 2 locales (often en + es or en + fr) prove the pipeline. Only then expand.
Source
prime-system/examples/frontend-design/primes/compiled/@community/feedback-i18n-launch-v1/atom.yaml