aicodingstack · ericyangpan · Jul 29, 2026 · Jul 29, 2026 · Jul 29, 2026
@@ -37,9 +37,8 @@ This is an evidence-review workflow with a deterministic importer. The Node scri
 ```bash
 node .agents/skills/benchmark-fetcher/scripts/fetch-benchmarks.mjs evidence.json
 node .agents/skills/benchmark-fetcher/scripts/fetch-benchmarks.mjs evidence.json --apply
-npm run changelog:generate
-npm run test:validate
-npm run data-health:check
+pnpm test:validate
+pnpm data-health:check
 ```
 
-Run `npm run check` and `npm run test:ci` before release handoff.
+Run `pnpm check` and `pnpm test:ci` before release handoff.
@@ -55,10 +55,10 @@ When translating:
 ## Validate
 
 ```bash
-npm run validate:i18n
-npm run validate:i18n-usage
-npm run validate:i18n-duplicates
-npm run test:validate
+pnpm validate:i18n
+pnpm validate:i18n-usage
+pnpm validate:i18n-duplicates
+pnpm test:validate
 ```
 
-For release-ready work, also run `npm run check` and `npm run test:ci`.
+For release-ready work, also run `pnpm check` and `pnpm test:ci`.
@@ -49,11 +49,10 @@ The merge helper is advisory. Always inspect its proposed result before applying
 At minimum run:
 
 ```bash
-npm run changelog:generate
-npm run generate
-npm run test:validate
-npm run validate:i18n
-npm run data-health:check
+pnpm generate
+pnpm test:validate
+pnpm validate:i18n
+pnpm data-health:check
 ```
 
-Run `npm run check` and `npm run test:ci` before handing off a release-ready change.
+Run `pnpm check` and `pnpm test:ci` before handing off a release-ready change.
@@ -166,11 +166,10 @@ console.log('📝 Required Validation')
 console.log('━'.repeat(60))
 console.log('')
 console.log('After editing and reviewing the manifest:')
-console.log('1. npm run changelog:generate')
-console.log('2. npm run generate')
-console.log('3. npm run test:validate')
-console.log('4. npm run validate:i18n')
-console.log('5. npm run data-health:check')
+console.log('1. pnpm generate')
+console.log('2. pnpm test:validate')
+console.log('3. pnpm validate:i18n')
+console.log('4. pnpm data-health:check')
 console.log('')
 console.log('━'.repeat(60))
 console.log('')

@@ -174,10 +174,10 @@ export class FieldTracker {
       output +=
         '2. Resolve required fields from authoritative sources; do not guess or add JSON comments\n'
       output += '3. Update provenance and verification metadata\n'
-      output += '4. Run validation: npm run test:validate\n'
+      output += '4. Run validation: pnpm test:validate\n'
     } else {
       output += '2. Update verified field if data confirmed accurate\n'
-      output += '3. Run validation: npm run test:validate\n'
+      output += '3. Run validation: pnpm test:validate\n'
     }
 
     return output

@@ -311,9 +311,9 @@ export function generateChangeReport(changes, mode, manifestPath) {
   output += `1. Review merged manifest: ${manifestPath}\n`
   if (summary.needsReview > 0) {
     output += '2. Resolve fields marked "Needs Review"\n'
-    output += '3. Run validation: npm run test:validate\n'
+    output += '3. Run validation: pnpm test:validate\n'
   } else {
-    output += '2. Run validation: npm run test:validate\n'
+    output += '2. Run validation: pnpm test:validate\n'
   }
 
   return output

@@ -172,7 +172,7 @@ The skill uses a 3-tier fallback strategy to map website model names to manifest
 
 3. **Validate**
    ```bash
-   npm run test:validate
+   pnpm test:validate
    ```
 
 4. **Commit Changes**

@@ -294,7 +294,7 @@ Average time per benchmark:   7.5s
 1. Review updated manifests in manifests/models/
 2. Add unmapped models to references/model-name-mappings.json
 3. Retry failed benchmarks if needed
-4. Run validation: npm run test:validate
+4. Run validation: pnpm test:validate
 5. Commit changes when satisfied
 ```
 
@@ -338,7 +338,7 @@ await mcp__chrome-devtools__take_screenshot({
 
 1. **Run during off-peak hours** to avoid rate limiting
 2. **Review unmapped models** and update mappings before next run
-3. **Validate manifests** after updates: `npm run test:validate`
+3. **Validate manifests** after updates: `pnpm test:validate`
 4. **Check for website changes** if extraction fails repeatedly
 5. **Keep mappings updated** as new models appear on leaderboards
 
@@ -380,7 +380,7 @@ Always validate manifests after updates:
 
 ```bash
 # Run schema validation
-npm run test:validate
+pnpm test:validate
 
 # Check JSON formatting
 node -c manifests/models/*.json
@@ -391,5 +391,5 @@ node -c manifests/models/*.json
 1. **Review updates**: Check manifest changes make sense
 2. **Update mappings**: Add newly discovered models to `model-name-mappings.json`
 3. **Retry failures**: Re-run with `--benchmarks` for failed benchmarks
-4. **Validate**: Run `npm run test:validate` to ensure schema compliance
+4. **Validate**: Run `pnpm test:validate` to ensure schema compliance
 5. **Commit changes**: Commit updated manifests to repository
@@ -234,7 +234,7 @@ function printNextSteps(isDryRun, unmappedModels, benchmarkResults) {
   }
 
   if (!isDryRun) {
-    steps.push('Run validation: npm run test:validate')
+    steps.push('Run validation: pnpm test:validate')
     steps.push('Commit changes when satisfied')
   }
 

@@ -88,7 +88,7 @@ node .claude/skills/manifest-automation/scripts/automate.mjs update extension co
    - Write manifest JSON to appropriate path
    - For IDE/CLI/Extension: check vendor manifest exists, create if missing
    - Update `data/github-stars.json`
-   - Run `npm run generate` to regenerate TypeScript data files
+   - Run `pnpm generate` to regenerate TypeScript data files
    - Run schema validation
    - Report success/failures to user
 
@@ -342,7 +342,7 @@ Status: DRAFT (3 fields incomplete)
 2. Manually fill TODO-marked fields if information available
 3. Add i18n translations (zh-Hans, de, ko)
 4. Update verified field once data confirmed accurate
-5. Run validation: npm run test:validate
+5. Run validation: pnpm test:validate
 
 Note: Vendor manifest, github-stars.json, and TypeScript files already updated.
 ```
@@ -433,7 +433,7 @@ After running this skill, expect:
 1. **Manifest file**: `manifests/<type>s/<name>.json`
 2. **With TODO comments**: For failed field extractions
 3. **Valid against schema**: `manifests/$schemas/<type>.schema.json`
-4. **Ready for validation**: Run `npm run test:validate`
+4. **Ready for validation**: Run `pnpm test:validate`
 
 ## GitHub Stars Update
 
@@ -520,7 +520,7 @@ updateGithubStarsEntry(manifestType, manifestName, { isNew: operationMode === 'c
 3. **Add translations**: Populate `i18n` object with localized content (zh-Hans, de, ko)
 4. **Set verified**: Change `verified` to `true` if data is confirmed accurate
 5. **Add related products**: Manually curate `relatedProducts` array
-6. **Run validation**: Ensure schema compliance with `npm run test:validate`
+6. **Run validation**: Ensure schema compliance with `pnpm test:validate`
 7. **Commit changes**: Add manifest to git repository
 
 Note: Vendor manifest and github-stars.json are automatically handled during creation.

@@ -173,10 +173,10 @@ export class FieldTracker {
     if (report.failed > 0) {
       output += '2. Manually fill TODO-marked fields if information available\n'
       output += '3. Update verified field once data confirmed accurate\n'
-      output += '4. Run validation: npm run test:validate\n'
+      output += '4. Run validation: pnpm test:validate\n'
     } else {
       output += '2. Update verified field if data confirmed accurate\n'
-      output += '3. Run validation: npm run test:validate\n'
+      output += '3. Run validation: pnpm test:validate\n'
     }
 
     return output

@@ -311,9 +311,9 @@ export function generateChangeReport(changes, mode, manifestPath) {
   output += `1. Review merged manifest: ${manifestPath}\n`
   if (summary.needsReview > 0) {
     output += '2. Resolve fields marked "Needs Review"\n'
-    output += '3. Run validation: npm run test:validate\n'
+    output += '3. Run validation: pnpm test:validate\n'
   } else {
-    output += '2. Run validation: npm run test:validate\n'
+    output += '2. Run validation: pnpm test:validate\n'
   }
 
   return output

@@ -269,7 +269,7 @@ This workflow guides you through creating or updating a CLI manifest using advan
 
 1. **Run schema validation**:
    ```bash
-   npm run test:validate
+   pnpm test:validate
    ```
 
 2. **Fix any validation errors**:

@@ -167,7 +167,7 @@ Follow **Phase 6** from CLI workflow:
 
 Run validation and fix errors:
 ```bash
-npm run test:validate
+pnpm test:validate
 ```
 
 ## Common Marketplace Patterns

@@ -209,7 +209,7 @@ Follow standard entity extraction:
 
 Run validation:
 ```bash
-npm run test:validate
+pnpm test:validate
 ```
 
 ## Common Patterns

@@ -142,7 +142,7 @@ Set to null if not applicable.
 
 Run validation:
 ```bash
-npm run test:validate
+pnpm test:validate
 ```
 
 ## Provider Type Examples

@@ -119,7 +119,7 @@ This workflow guides you through creating or updating a Vendor manifest - the si
 
 Run validation:
 ```bash
-npm run test:validate
+pnpm test:validate
 ```
 
 ## Examples

@@ -0,0 +1,19 @@
+name: Setup pnpm
+description: Install the pinned pnpm and restore project dependencies
+
+runs:
+  using: composite
+  steps:
+    - name: Setup pnpm
+      uses: pnpm/action-setup@v6
+
+    - name: Setup Node.js
+      uses: actions/setup-node@v7
+      with:
+        node-version: '22'
+        cache: pnpm
+        cache-dependency-path: pnpm-lock.yaml
+
+    - name: Install dependencies
+      shell: bash
+      run: pnpm install --frozen-lockfile
@@ -1,6 +1,6 @@
 version: 2
 updates:
-  # npm dependencies
+  # JavaScript dependencies (GitHub names this ecosystem "npm" for npm-compatible registries)
   - package-ecosystem: "npm"
     directory: "/"
     schedule:

@@ -54,11 +54,11 @@ Closes #(issue number)
 
 <!-- Describe how you tested your changes -->
 
-- [ ] Local build passes (`npm run build`)
-- [ ] All validation checks pass (`npm run test:validate`)
-- [ ] URL validation passes (`npm run validate:urls`)
-- [ ] Static analysis passes (`npm run biome:check`)
-- [ ] Spell check passes (`npm run spell`)
+- [ ] Local build passes (`pnpm build`)
+- [ ] All validation checks pass (`pnpm test:validate`)
+- [ ] URL validation passes (`pnpm validate:urls`)
+- [ ] Static analysis passes (`pnpm biome:check`)
+- [ ] Spell check passes (`pnpm spell`)
 
 ## Screenshots (if applicable)
 

@@ -21,20 +21,14 @@ jobs:
       - name: Checkout repository
         uses: actions/checkout@v7
 
-      - name: Setup Node.js
-        uses: actions/setup-node@v7
-        with:
-          node-version: '22'
-          cache: 'npm'
-
-      - name: Install dependencies
-        run: npm ci
+      - name: Setup pnpm
+        uses: ./.github/actions/setup-pnpm
 
       - name: Check current data health
         id: data-health
         continue-on-error: true
         run: |
-          npx tsx scripts/validate/data-health.ts \
+          pnpm exec tsx scripts/validate/data-health.ts \
             --as-of="$(TZ=Asia/Shanghai date +%F)" \
             --fail-on=error \
             --fail-on-code=stale-verification,future-verification-date \