chore(prep-prod): CI + SEO + observabilite + feature flag + i18n P26

.env.example

@@ -1,2 +1,16 @@
-# URL interne du backend (server-side only)
+# URL interne du backend (server-side only) — utilise dans hooks.server.ts
 API_URL=http://localhost:3001/api
+
+# URL exposee au client (visible dans le bundle) — utilise pour SSR fetch client + Playwright
+PUBLIC_API_BASE_URL=http://localhost:3001
+
+# Feature flags — voir src/lib/config/features.ts
+# P26 v2 workflow challenge (slices, validations, digest maintainers).
+# OFF par defaut ; passer a 'true' quand le back livre SKI-72..91, 115..123.
+PUBLIC_P26_ENABLED=false
+
+# Observabilite (optionnelle) — voir src/lib/observability.ts
+# PUBLIC_SENTRY_DSN=
+# PUBLIC_APP_VERSION=
+# PUBLIC_POSTHOG_KEY=
+# PUBLIC_POSTHOG_HOST=https://eu.i.posthog.com

.github/workflows/ci.yml

@@ -99,8 +99,11 @@ jobs:
         run: npm run build
 
       - name: Run Lighthouse CI
+        # Assertion a11y = "error" a 100 -> ce job BLOQUE la PR sur regression
+        # a11y. Perf/best/seo restent en warn (variance mobile trop haute pour
+        # bloquer sans false positives).
         run: |
           npm install -g @lhci/cli@0.15.x
-          lhci autorun --config=./lighthouserc.json || true
+          lhci autorun --config=./lighthouserc.json
         env:
           LHCI_GITHUB_APP_TOKEN: ${{ secrets.LHCI_GITHUB_APP_TOKEN }}

.gitignore

@@ -36,7 +36,8 @@ Thumbs.db
 
 
 # Documentation
-docs/
+docs/*
+!docs/LIGHTHOUSE-AUDIT-*.md
 
 # BMAD workflow
 _bmad/
@@ -47,4 +48,13 @@ design-artifacts/
 
 # AI coding assistant caches
 .claude/
-.claude.*
+.claude.*
+# Lighthouse audit output
+lighthouse-report/
+
+# Axe audit output
+axe-report/
+
+# Bundle analyzer output
+bundle-report.html
+stats.html

docs/LIGHTHOUSE-AUDIT-2026-08-10.md

@@ -0,0 +1,45 @@
+# Lighthouse mobile audit — 2026-08-10
+
+Baseline mobile 4G throttled sur build production (`npm run preview`).
+
+## Scores apres fixes perf
+
+Fixes appliques dans le commit `chore(perf): CLS fallback fonts + img dimensions + preconnect API` :
+
+- Fallback fonts synthetiques (Fraunces Fallback / Bricolage Fallback) avec `size-adjust` + `ascent-override` + `descent-override` calibres empiriquement — elimine le layout shift au font swap.
+- `<link rel="preconnect" href="https://api.skill-uv.com" crossorigin />` dans `app.html` — coupe le cout DNS + TLS des premiers fetch API cote client.
+- 16 tags `<img>` : dimensions explicites `width` + `height` derivees des classes Tailwind parentes + `loading="lazy"` sur 7 below-the-fold.
+
+| Page                  | Perf | A11y | Best-practices | SEO |
+| --------------------- | ---: | ---: | -------------: | --: |
+| `/`                   |   74 |  100 |            100 | 100 |
+| `/challenges`         |   62 |  100 |             96 | 100 |
+| `/pricing`            |   77 |   95 |             96 | 100 |
+| `/for-maintainers`    |   85 |   96 |             96 | 100 |
+| `/verify/{hash}`      |   75 |   95 |            100 | 100 |
+
+Rerun : `npm run audit:lighthouse` (dev server) ou `LH_BASE_URL=http://localhost:4173 npm run audit:lighthouse` (preview).
+
+## Baseline pre-fixes (pour reference)
+
+| Page                  | Perf | A11y | Best | SEO |
+| --------------------- | ---: | ---: | ---: | --: |
+| `/`                   |   74 |  100 |  100 | 100 |
+| `/challenges`         |   79 |  100 |   96 | 100 |
+| `/pricing`            |   76 |   95 |   96 | 100 |
+| `/for-maintainers`    |   76 |   96 |   96 | 100 |
+| `/verify/{hash}`      |   76 |   95 |  100 | 100 |
+
+## Analyse
+
+- **`/for-maintainers` +9** — gain net attribuable au CLS (page riche en badges SVG et images shields-like, qui reflowaient sans dimensions).
+- **`/pricing` stable** — page dense en cards, deja optimisee.
+- **`/`, `/verify/{hash}` stables** — perf limitee par le bundle JS initial, pas les CLS.
+- **`/challenges` -17** — regression a investiguer. Hypotheses : la variabilite Lighthouse mobile est de +/-15 points par run (throttling CPU aleatoire). Le score baseline avait ete pris sur un run isole. Un run stable donnerait probablement 70-80.
+
+## Prochaines etapes suggerees (post-launch)
+
+1. `npm i --save-dev vite-bundle-visualizer` puis `vite build --report` pour voir la carte du bundle. Cibler split routes lourdes (monaco-editor, gsap).
+2. Audit `@axe-core/cli` pour les 4-5% a11y manquants sur `/pricing`, `/for-maintainers`, `/verify/{hash}` (probablement contrast couleurs badges + `aria-label` icones lucide).
+3. Lighthouse CI (`.github/workflows/lighthouse.yml`) pour tracker les regressions en CI plutot que de faire des audits ponctuels.
+4. Considerer `image-optim` sur les SVG static et convertir les PNG en WebP (favicon-192, icon-512).

lighthouserc.json

@@ -6,22 +6,25 @@
 			"url": [
 				"http://localhost:4173/",
 				"http://localhost:4173/challenges",
-				"http://localhost:4173/pricing"
+				"http://localhost:4173/pricing",
+				"http://localhost:4173/for-maintainers",
+				"http://localhost:4173/verify/000000000000000000000000000000000000000000000000000000000deadbeef"
 			],
-			"numberOfRuns": 1,
+			"numberOfRuns": 3,
 			"settings": {
 				"preset": "desktop",
 				"emulatedFormFactor": "mobile",
 				"throttlingMethod": "simulate",
-				"onlyCategories": ["performance", "accessibility", "best-practices", "seo"]
+				"onlyCategories": ["performance", "accessibility", "best-practices", "seo"],
+				"skipAudits": ["uses-http2"]
 			}
 		},
 		"assert": {
 			"assertions": {
 				"categories:performance": ["warn", { "minScore": 0.7 }],
-				"categories:accessibility": ["error", { "minScore": 0.9 }],
-				"categories:best-practices": ["warn", { "minScore": 0.85 }],
-				"categories:seo": ["warn", { "minScore": 0.85 }]
+				"categories:accessibility": ["error", { "minScore": 1.0 }],
+				"categories:best-practices": ["warn", { "minScore": 0.9 }],
+				"categories:seo": ["warn", { "minScore": 0.9 }]
 			}
 		},
 		"upload": {