Fix: Update comment moderation logic (#38451)

# Fix: Update comment moderation logic

Inspired by #38450 to help fix/complete the moderation script,
this updates the moderation script to:

- Add language detection for non-English comments
- Implement profanity filtering
- Check for off-topic content
- Improve forbidden link detection
- Add owner exemption logic
- Enhance error handling and logging
- Add a message indicating why the comment was deleted.

*To configure the moderation lists:*

- Go to repository Settings > Secrets > Actions
- Add `MODERATION_PROFANITY` secret containing a JSON array of terms
- Add `MODERATION_OFFTOPIC` secret containing a JSON array of phrases
- Example format for each: `["term1","term2","term3"]`

Resources for priming the PROFANITY and OFFTOPIC lists:
- https://www.cs.cmu.edu/~biglou/resources/ (first link)
- https://github.com/OOPSpam/spam-words/blob/main/spam-words-EN.txt

Co-authored-by: Laurent Destailleur <eldy@destailleur.fr>

.github/workflows/moderation.yml

@@ -4,7 +4,6 @@ name: Moderation Bot
 on:
   issue_comment:
     types: [created, edited]
-
   pull_request_review_comment:
     types: [created, edited]
 
@@ -15,45 +14,63 @@ permissions:
 jobs:
   moderate:
     runs-on: ubuntu-latest
-
     steps:
+      - name: Install franc
+        run: npm install franc
+
       - name: Moderate comments
         uses: actions/github-script@v9
         with:
           github-token: ${{ secrets.GITHUB_TOKEN }}
           script: |
-            const franc = require('franc')
-
-            const comment = context.payload.comment.body || ''
-            const lower = comment.toLowerCase()
-
-            // Détection YouTube Music
             const forbiddenPatterns = [
               'music.youtube.com',
               'youtu.be',
               'youtube.com/watch'
             ]
 
-            const hasYoutubeMusic = forbiddenPatterns.some(p => lower.includes(p))
+            const profanityList = JSON.parse(process.env.PROFANITY_LIST || '[]')
+            const offTopicKeywords = JSON.parse(process.env.OFFTOPIC_LIST || '[]')
+
+            const comment = context.payload.comment.body || ''
+            const commentAuthor = context.payload.comment.user.login
+            const repoOwner = context.repo.owner
+            const isOwner = commentAuthor === repoOwner
+
+            const isForbidden = forbiddenPatterns.some(p => comment.toLowerCase().includes(p))
+            const isProfane = profanityList.some(word => new RegExp(`\\b${word}\\b`, 'i').test(comment))
+            const isOffTopic = offTopicKeywords.some(kw => comment.toLowerCase().includes(kw.toLowerCase()))
 
-            // Détection langue
             let isEnglish = true
-
             try {
+              const franc = require('franc')
               const lang = franc(comment)
-              isEnglish = lang === 'eng'
+              isEnglish = lang && lang.startsWith('eng')
             } catch (e) {
               console.log('Language detection failed:', e)
             }
 
-            console.log('Detected language:', isEnglish ? 'english' : 'not english')
+            if ((isForbidden || isProfane || isOffTopic || !isEnglish) && !isOwner) {
+              try {
+                const issueNumber = context.payload.issue?.number || context.payload.pull_request?.number
+                const reason = !isEnglish ? 'non-English comment' :
+                              isForbidden ? 'forbidden link' :
+                              isProfane ? 'profanity' : 'off-topic content'
 
-            if (hasYoutubeMusic) {
-              console.log('Deleting comment...')
+                await github.rest.issues.createComment({
+                  owner: context.repo.owner,
+                  repo: context.repo.repo,
+                  issue_number: issueNumber,
+                  body: `Comment auto removed: ${reason} violates repository guidelines.`
+                })
 
                 await github.rest.issues.deleteComment({
                   owner: context.repo.owner,
                   repo: context.repo.repo,
                   comment_id: context.payload.comment.id
                 })
+              } catch (error) {
+                console.error('Moderation error:', error)
+                throw error
+              }
             }