Detect ai spam posts and remove them.

[ampelectrecuted]

I don't want to use another AI for this as omniblocks alone probably emits a bunch of carbon dioxide from just all the AI github actions. Instead we can check patterns using regex:

• Skip [bot] accounts for obvious reasons.
• Em dashes— a huge pattern of AI generated text— is automatically flagged. Nobody in a GitHub comments section would use em dashes unless they were just clankers that Microslop (often misspelt as Microsoft) wants you to use.
• Words like "robust", "deep dive into", etc. = flagged.
• turn0search0 is flagged.
• If your issue starts with "Subject:" or "Title:" it is flagged.
• Emojis are flagged.

If at least 3 patterns are flagged then your comment is hidden, and you will be pinged with a message saying

Please do not post AI text on this issue tracker. If you want to talk to AI ping @ coderabbitai or @ OmniBlocks/ai. To generate PR with @ cto-new just ping that user.
(remove te spaces after the ats)

[traycerai]

Plan

Observations

The repository already has a spam detection workflow at file:.github/workflows/spam-guard.yml that currently uses AI (Ministral-3B model via actions/ai-inference@v1) to detect spam from first-time commenters. The workflow already filters out bot accounts and has mechanisms to minimize comments and post notifications. The existing structure includes first-time contributor checks, comment minimization via GraphQL API, and audit logging.

Approach

Replace the AI-based spam detection in file:.github/workflows/spam-guard.yml with regex-based pattern matching to detect AI-generated text. The workflow will check for specific patterns (em dashes, AI buzzwords like "robust" and "deep dive into", "turn0search0", formal headers, and emojis). When 3 or more patterns are detected, the comment will be minimized and the user will receive a notification directing them to appropriate AI tools. This approach eliminates AI inference costs while maintaining spam protection.

Implementation Steps

1. Modify Spam Detection Logic in file:.github/workflows/spam-guard.yml

Remove AI-based detection steps:

• Delete step "3️⃣ Analyze comment for spam with AI" (lines 78-125) that uses actions/ai-inference@v1
• Delete step "4️⃣ Parse AI classification" (lines 129-146)

Add regex-based pattern detection step:

• Insert a new step after "2️⃣ Skip spam check for known contributors" (after line 73)
• Name it "3️⃣ Check for AI-generated text patterns"
• Use bash script to analyze ${{ github.event.comment.body }}
• Implement pattern checks:
  • Em dashes: Regex — (Unicode U+2014)
  • AI buzzwords: Case-insensitive regex for words like robust, deep dive into, leverage, utilize, comprehensive, holistic, synergy, paradigm
  • turn0search0: Exact match for turn0search0
  • Formal headers: Regex ^(Subject|Title): at start of comment
  • Emojis: Unicode emoji ranges (U+1F600-U+1F64F, U+1F300-U+1F5FF, U+1F680-U+1F6FF, U+2600-U+26FF, U+2700-U+27BF)
• Count flagged patterns and output to $GITHUB_OUTPUT as pattern_count and is_ai_spam (true if count >= 3)

2. Update Comment Handling Logic

Modify step "5️⃣ Handle SPAM comment":

• Change condition from steps.classification.outputs.verdict == 'SPAM' to steps.pattern-check.outputs.is_ai_spam == 'true'
• Instead of deleting the comment, minimize it using GraphQL mutation (similar to SUSPICIOUS handling in lines 202-207)
• Replace the notification message with the required text:

  > Please do not post AI text on this issue tracker. If you want to talk to AI ping @coderabbitai or @OmniBlocks/ai. To generate PR with @cto-new just ping that user.
  

• Post this as a reply to the flagged comment (tag the user with @${{ github.event.sender.login }})
• Remove maintainer notification (no cc to @supervoidcoder)

Remove or repurpose step "6️⃣ Handle SUSPICIOUS comment":

• Since the new logic only has binary classification (AI spam or not), remove the SUSPICIOUS handling
• Alternatively, keep it for future use but it won't be triggered by the new pattern-based detection

Update step "7️⃣ Log SAFE comments":

• Change condition from steps.classification.outputs.verdict == 'SAFE' to steps.pattern-check.outputs.is_ai_spam != 'true'

3. Update Audit Logging

Modify step "8️⃣ Add comprehensive audit log":

• Update the audit log format to include pattern count instead of AI verdict
• Change from verdict=${{ steps.classification.outputs.verdict }} to pattern_count=${{ steps.pattern-check.outputs.pattern_count }} is_ai_spam=${{ steps.pattern-check.outputs.is_ai_spam }}

4. Ensure Bot Account Filtering

Verify existing bot filtering:

• The workflow already has if: github.event.sender.type == 'User' at line 18
• This correctly skips [bot] accounts as required
• No changes needed for this requirement

5. Testing Considerations

Pattern detection validation:

• Test with comments containing em dashes: "This is great—really impressive work—thanks!"
• Test with AI buzzwords: "This is a robust solution that provides a deep dive into the comprehensive architecture"
• Test with turn0search0: "turn0search0 shows that..."
• Test with formal headers: "Subject: Bug Report\nTitle: Issue with feature"
• Test with emojis: "Great work! 🎉 This is amazing! 👍"
• Test with mixed patterns to verify 3+ threshold
• Test with legitimate comments to ensure no false positives

Edge cases:

• Comments with 1-2 patterns should not be flagged
• Bot accounts should be completely skipped
• Known contributors should bypass all checks (existing logic)

6. Documentation Updates

Update workflow comments:

• Modify the comment at line 76 from "Analyze comment with AI for spam/phishing" to "Check for AI-generated text patterns using regex"
• Update inline documentation to reflect regex-based approach instead of AI classification

Pattern Detection Implementation Details

The bash script should implement pattern matching as follows:

COMMENT_BODY="${{ github.event.comment.body }}"
PATTERN_COUNT=0

# Check for em dashes
if echo "$COMMENT_BODY" | grep -q '—'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for AI buzzwords (case-insensitive)
if echo "$COMMENT_BODY" | grep -qiE '(robust|deep dive into|leverage|utilize|comprehensive|holistic|synergy|paradigm)'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for turn0search0
if echo "$COMMENT_BODY" | grep -q 'turn0search0'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for formal headers at start
if echo "$COMMENT_BODY" | grep -qE '^(Subject|Title):'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for emojis (Unicode ranges)
if echo "$COMMENT_BODY" | grep -qP '[\x{1F600}-\x{1F64F}\x{1F300}-\x{1F5FF}\x{1F680}-\x{1F6FF}\x{2600}-\x{26FF}\x{2700}-\x{27BF}]'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Output results
echo "pattern_count=$PATTERN_COUNT" >> "$GITHUB_OUTPUT"
if [ "$PATTERN_COUNT" -ge 3 ]; then
  echo "is_ai_spam=true" >> "$GITHUB_OUTPUT"
else
  echo "is_ai_spam=false" >> "$GITHUB_OUTPUT"
fi

Workflow Sequence Diagram

sequenceDiagram
    participant User
    participant GitHub
    participant Workflow
    participant GraphQL

    User->>GitHub: Posts comment on issue
    GitHub->>Workflow: Triggers issue_comment event
    
    alt Sender is Bot
        Workflow->>Workflow: Skip (sender.type != 'User')
    else Sender is User
        Workflow->>GitHub: Check contribution history
        
        alt Known Contributor
            Workflow->>Workflow: Skip spam check
        else First-time Commenter
            Workflow->>Workflow: Run regex pattern checks
            Workflow->>Workflow: Count flagged patterns
            
            alt Pattern count >= 3
                Workflow->>GraphQL: Minimize comment (OFF_TOPIC)
                Workflow->>GitHub: Post notification to user
                Workflow->>GitHub: Add audit log
            else Pattern count < 3
                Workflow->>Workflow: Log as safe
                Workflow->>GitHub: Add audit log
            end
        end
    end

Loading

Import In IDE

🤖 Prompt for AI Agents

## Observations

The repository already has a spam detection workflow at `file:.github/workflows/spam-guard.yml` that currently uses AI (Ministral-3B model via `actions/ai-inference@v1`) to detect spam from first-time commenters. The workflow already filters out bot accounts and has mechanisms to minimize comments and post notifications. The existing structure includes first-time contributor checks, comment minimization via GraphQL API, and audit logging.

## Approach

Replace the AI-based spam detection in `file:.github/workflows/spam-guard.yml` with regex-based pattern matching to detect AI-generated text. The workflow will check for specific patterns (em dashes, AI buzzwords like "robust" and "deep dive into", "turn0search0", formal headers, and emojis). When 3 or more patterns are detected, the comment will be minimized and the user will receive a notification directing them to appropriate AI tools. This approach eliminates AI inference costs while maintaining spam protection.

## Implementation Steps

### 1. Modify Spam Detection Logic in `file:.github/workflows/spam-guard.yml`

**Remove AI-based detection steps:**
- Delete step "3️⃣ Analyze comment for spam with AI" (lines 78-125) that uses `actions/ai-inference@v1`
- Delete step "4️⃣ Parse AI classification" (lines 129-146)

**Add regex-based pattern detection step:**
- Insert a new step after "2️⃣ Skip spam check for known contributors" (after line 73)
- Name it "3️⃣ Check for AI-generated text patterns"
- Use bash script to analyze `${{ github.event.comment.body }}`
- Implement pattern checks:
  - **Em dashes**: Regex `—` (Unicode U+2014)
  - **AI buzzwords**: Case-insensitive regex for words like `robust`, `deep dive into`, `leverage`, `utilize`, `comprehensive`, `holistic`, `synergy`, `paradigm`
  - **turn0search0**: Exact match for `turn0search0`
  - **Formal headers**: Regex `^(Subject|Title):` at start of comment
  - **Emojis**: Unicode emoji ranges (U+1F600-U+1F64F, U+1F300-U+1F5FF, U+1F680-U+1F6FF, U+2600-U+26FF, U+2700-U+27BF)
- Count flagged patterns and output to `$GITHUB_OUTPUT` as `pattern_count` and `is_ai_spam` (true if count >= 3)

### 2. Update Comment Handling Logic

**Modify step "5️⃣ Handle SPAM comment":**
- Change condition from `steps.classification.outputs.verdict == 'SPAM'` to `steps.pattern-check.outputs.is_ai_spam == 'true'`
- Instead of deleting the comment, minimize it using GraphQL mutation (similar to SUSPICIOUS handling in lines 202-207)
- Replace the notification message with the required text:
  ```
  > Please do not post AI text on this issue tracker. If you want to talk to AI ping @coderabbitai or @OmniBlocks/ai. To generate PR with @cto-new just ping that user.
  ```
- Post this as a reply to the flagged comment (tag the user with `@${{ github.event.sender.login }}`)
- Remove maintainer notification (no cc to @supervoidcoder)

**Remove or repurpose step "6️⃣ Handle SUSPICIOUS comment":**
- Since the new logic only has binary classification (AI spam or not), remove the SUSPICIOUS handling
- Alternatively, keep it for future use but it won't be triggered by the new pattern-based detection

**Update step "7️⃣ Log SAFE comments":**
- Change condition from `steps.classification.outputs.verdict == 'SAFE'` to `steps.pattern-check.outputs.is_ai_spam != 'true'`

### 3. Update Audit Logging

**Modify step "8️⃣ Add comprehensive audit log":**
- Update the audit log format to include pattern count instead of AI verdict
- Change from `verdict=${{ steps.classification.outputs.verdict }}` to `pattern_count=${{ steps.pattern-check.outputs.pattern_count }} is_ai_spam=${{ steps.pattern-check.outputs.is_ai_spam }}`

### 4. Ensure Bot Account Filtering

**Verify existing bot filtering:**
- The workflow already has `if: github.event.sender.type == 'User'` at line 18
- This correctly skips `[bot]` accounts as required
- No changes needed for this requirement

### 5. Testing Considerations

**Pattern detection validation:**
- Test with comments containing em dashes: "This is great—really impressive work—thanks!"
- Test with AI buzzwords: "This is a robust solution that provides a deep dive into the comprehensive architecture"
- Test with turn0search0: "turn0search0 shows that..."
- Test with formal headers: "Subject: Bug Report\nTitle: Issue with feature"
- Test with emojis: "Great work! 🎉 This is amazing! 👍"
- Test with mixed patterns to verify 3+ threshold
- Test with legitimate comments to ensure no false positives

**Edge cases:**
- Comments with 1-2 patterns should not be flagged
- Bot accounts should be completely skipped
- Known contributors should bypass all checks (existing logic)

### 6. Documentation Updates

**Update workflow comments:**
- Modify the comment at line 76 from "Analyze comment with AI for spam/phishing" to "Check for AI-generated text patterns using regex"
- Update inline documentation to reflect regex-based approach instead of AI classification

## Pattern Detection Implementation Details

The bash script should implement pattern matching as follows:

```bash
COMMENT_BODY="${{ github.event.comment.body }}"
PATTERN_COUNT=0

# Check for em dashes
if echo "$COMMENT_BODY" | grep -q '—'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for AI buzzwords (case-insensitive)
if echo "$COMMENT_BODY" | grep -qiE '(robust|deep dive into|leverage|utilize|comprehensive|holistic|synergy|paradigm)'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for turn0search0
if echo "$COMMENT_BODY" | grep -q 'turn0search0'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for formal headers at start
if echo "$COMMENT_BODY" | grep -qE '^(Subject|Title):'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Check for emojis (Unicode ranges)
if echo "$COMMENT_BODY" | grep -qP '[\x{1F600}-\x{1F64F}\x{1F300}-\x{1F5FF}\x{1F680}-\x{1F6FF}\x{2600}-\x{26FF}\x{2700}-\x{27BF}]'; then
  PATTERN_COUNT=$((PATTERN_COUNT + 1))
fi

# Output results
echo "pattern_count=$PATTERN_COUNT" >> "$GITHUB_OUTPUT"
if [ "$PATTERN_COUNT" -ge 3 ]; then
  echo "is_ai_spam=true" >> "$GITHUB_OUTPUT"
else
  echo "is_ai_spam=false" >> "$GITHUB_OUTPUT"
fi
```

## Workflow Sequence Diagram

```mermaid
sequenceDiagram
    participant User
    participant GitHub
    participant Workflow
    participant GraphQL

    User->>GitHub: Posts comment on issue
    GitHub->>Workflow: Triggers issue_comment event
    
    alt Sender is Bot
        Workflow->>Workflow: Skip (sender.type != 'User')
    else Sender is User
        Workflow->>GitHub: Check contribution history
        
        alt Known Contributor
            Workflow->>Workflow: Skip spam check
        else First-time Commenter
            Workflow->>Workflow: Run regex pattern checks
            Workflow->>Workflow: Count flagged patterns
            
            alt Pattern count >= 3
                Workflow->>GraphQL: Minimize comment (OFF_TOPIC)
                Workflow->>GitHub: Post notification to user
                Workflow->>GitHub: Add audit log
            else Pattern count < 3
                Workflow->>Workflow: Log as safe
                Workflow->>GitHub: Add audit log
            end
        end
    end
```

---

Execution Information

Branch: main
Commit: 2c93b63

---

💡 Tips

Supported Commands (Inside Comments)

• Use @traycerai generate to iterate on the previous version of the implementation plan.

Supported Commands (Inside Description)

• Add @traycerai ignore anywhere in the ticket description to prevent this ticket from being processed.
• Add @traycerai branch:<branch-name> anywhere in the ticket description to specify the target branch for the implementation plan.

Community

• Join our Discord Community to get help, request features, and share feedback.
• Follow us on X/Twitter for updates and announcements.

[github-actions]

Hi @ampelectrecuted! Thanks for bringing this up! Your idea to detect and filter potential AI-generated spam posts using patterns like regex is intriguing and could definitely help prevent clutter in the issue tracker. Flagging text based on formatting, language choices, or certain bot indicators seems like a strong approach for automation, but it might need careful tuning to avoid unintentionally flagging legitimate posts.

Your outlined criteria, such as skipping bot accounts, detecting unusual usage of em dashes, or specific phrases, are clever. However, flagging emojis might be too strict, as some users genuinely use them to express tone or illustrate ideas. If implemented, testing this system thoroughly would be essential to ensure it doesn't overly penalize users for harmless patterns.

If you want further advice or brainstorming on this, let me know! This could be a productive discussion for improving moderation tools.

[github-actions]

🎭 Issue Poem

In the land of OmniBlocks, where code takes flight,
We’ve got to dodge the bots that litter our site.
With regex on our side, we’ll scan and we’ll sift,
To catch AI spam and give it a swift lift!

Em dashes and "robust," we’ll say “not today!”
Let’s keep our comments clear, in a playful way.
So if patterns abound, and your text goes awry,
We'll gently remind you, “Let’s keep it real, bye-bye!” 🌟

---

Generated with ❤️ by the OmniBlocks Poetry Bot

[ampelectrecuted]

@coderabbitai is this a good idea and how to improve

[coderabbitai]

Tip

For best results, initiate chat on the files or code changes.

@ampelectrecuted The core idea is solid—using pattern matching instead of AI is more efficient and reduces resource usage. However, some patterns need refinement to avoid false positives:

Good aspects ✅

• Efficient approach: Regex-based detection is fast and doesn't require AI
• Threshold system: Requiring 3+ patterns reduces false positives
• Clear user guidance: The notification message directs users appropriately
• Bot exclusion: Skipping bot accounts makes sense

Concerns & Improvements 🔧

High false-positive patterns:

1. Emojis 🚨 - Many legitimate users use emojis. Consider removing this entirely or only flagging if combined with excessive emoji use (5+ in one comment)
2. Em dashes - Used in legitimate formal writing. Consider only flagging if there are 2+ in a single comment
3. Words like "robust" - Real developers use these too. Needs a larger, more specific word list

Suggested improvements:

Better AI-specific patterns:

- "delve into", "utilize", "leverage" (instead of use)
- "It's worth noting that", "It's important to mention"
- Overly apologetic: "I apologize for any confusion"
- Multiple greeting/closing pairs in short text
- "As a [role], I..."
- Bullet points with overly structured intro phrases

Additional safeguards:

• Account age check: New accounts (< 7 days) with 2 patterns could be flagged
• Comment length: Very short comments (< 100 chars) with patterns are less suspicious
• Whitelist: Allow org members/frequent contributors to bypass checks
• Logging: Track flagged comments for pattern tuning

Scoring system example:

• Em dash (2+ occurrences): 1 point
• AI buzzword: 1 point each (max 2 from this category)
• turn0search0: 2 points
• Subject:/Title: prefix: 2 points
• Excessive emojis (5+): 1 point
• Threshold: 4+ points = flag

Would you like me to create a sample implementation with these improvements?

---

🧠 Learnings used

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-12-05T16:55:31.105Z
Learning: OmniBlocks/scratch-gui: supervoidcoder wants automated comment moderation to detect spam/phishing from first-time commenters using GitHub Models AI (similar to ulti.yml workflow). Should analyze comments and hide/delete based on severity levels.

---

Learnt from: someCatInTheWorld
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2026-01-28T23:50:49.831Z
Learning: OmniBlocks/scratch-gui moderation: the spam rule currently flags the exact pattern “coderabbitai is this true”, causing comment jobs to fail, while “grok is this true” is not caught. Recommend avoiding that exact phrasing or adding punctuation/line breaks; consider relaxing the rule to reduce false positives.

---

Learnt from: github-actions[bot]
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:58:09.061Z
Learning: Extend spam-cleanup patterns in OmniBlocks/scratch-gui to include “auto-generated comment: … by coderabbit.ai” and structural markers (tips_start, analysis chain, agentComponentType, rawResChunk) to fully minimize CodeRabbit auto-noise.

---

Learnt from: github-actions[bot]
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:58:09.061Z
Learning: In OmniBlocks/scratch-gui, when cleaning PR threads, prefer minimizing CodeRabbit auto-status comments that contain phrases like “For best results, initiate chat…”, “Rate Limit Exceeded”, and “This is an automated status update…”, while leaving human-directed replies intact.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-10-10T15:23:20.794Z
Learning: OmniBlocks/scratch-gui PR test workflow now ignores bot/AI comments containing the trigger phrase; only human users are allowed to trigger the deployment.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:34:24.309Z
Learning: CodeRabbit experienced a spam bug on PR `#309` in OmniBlocks/scratch-gui where it posted 11 duplicate "no user input to reply to" messages. The spam was triggered by bot comments from github-actions and wellcode-ai. The comments were successfully minimized using GitHub GraphQL API mutations with the OFF_TOPIC classifier.

---

Learnt from: 8to16
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2026-01-22T19:24:58.994Z
Learning: OmniBlocks/scratch-gui moderation preference: When CodeRabbit detects a human comment with strong profanity or a clear Code of Conduct violation (e.g., slurs, harassment, threats, sexual content, doxxing), it must ping 8to16 and supervoidcoder in the reply. Do not ping for mild words like "crap" or "heck" (explicitly allowed by project policy). Do not trigger on bot/CI comments. Keep the tone calm and optionally include a brief CoC reminder. This rule applies only within the OmniBlocks/scratch-gui repository.

---

Learnt from: github-actions[bot]
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:50:28.065Z
Learning: OmniBlocks/scratch-gui: When cleaning PR noise, prefer minimizing CodeRabbit auto-generated comments matching headers like "This is an auto-generated reply by CodeRabbit", "Rate Limit Exceeded", and "automated status update", while preserving human-directed conversation.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:48:23.435Z
Learning: OmniBlocks/scratch-gui: For CI comment hygiene, prefer not mentioning bot/app requesters in the Mega PR Test & Analysis sticky comment; only mention humans. Track such behavior changes via a GitHub issue assigned to supervoidcoder with backlinks to the initiating PR and comment.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:55:33.086Z
Learning: OmniBlocks/scratch-gui: Never reply to CodeRabbit’s own auto-generated comments (status, tips, rate-limit); when a bot loop is detected on a PR, post a single apology, minimize CodeRabbit auto-noise, and then remain silent unless a human asks for more.

[ampelectrecuted]

also if your post contains stuff like "as of January 2022" or "as a large language model" it will be immediately marked as ai @coderabbitai

[coderabbitai]

Tip

For best results, initiate chat on the files or code changes.

@ampelectrecuted Excellent additions! Those are extremely strong AI indicators. In fact, they're so reliable they could warrant immediate flagging even on their own.

Strong AI self-identification patterns:

Cutoff date references:

as of (January|February|March|...) \d{4}
knowledge cutoff
my training data
last updated in

AI self-reference phrases:

as a large language model
as an AI( assistant)?
I('m| am) an AI
I don't have (personal|real-time|the ability)
I cannot (browse|access|verify) (the internet|real-time|external)

Scoring recommendation:

These patterns are so definitive that I'd suggest:

• Instant flag (single pattern = hide):

  • "as a large language model"
  • "as an AI"
  • "I'm an AI"

• High weight (2+ points each):

  • Cutoff date references ("as of January 2022")
  • "I cannot access the internet"
  • "my training data"
  • "turn0search0"
  • "Subject:"/Title:" prefix

• Medium weight (1 point):

  • Multiple em dashes
  • AI buzzwords (robust, leverage, etc.)
  • Excessive emojis (5+)

This way, even if someone tries to disguise AI text by mixing styles, the high-weight patterns will catch it, while avoiding false positives on legitimate comments that might just use one medium-weight pattern.

Would you like me to draft a complete pattern list with the scoring system?

---

✏️ Learnings added

Learnt from: ampelectrecuted
URL: https://github.com/OmniBlocks/scratch-gui/issues/768

Timestamp: 2026-02-04T17:16:22.593Z
Learning: In OmniBlocks/scratch-gui AI spam detection: Phrases like "as a large language model", "as an AI", "I'm an AI", or cutoff date references like "as of January 2022" are extremely strong indicators of AI-generated text and should trigger immediate flagging (single pattern sufficient).

Note: Learnings are effective only in the context of similar code segments. To apply general coding guidelines, please configure review instructions. You can manage existing learnings in the app.

---

🧠 Learnings used

Learnt from: someCatInTheWorld
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2026-01-28T23:50:49.831Z
Learning: OmniBlocks/scratch-gui moderation: the spam rule currently flags the exact pattern “coderabbitai is this true”, causing comment jobs to fail, while “grok is this true” is not caught. Recommend avoiding that exact phrasing or adding punctuation/line breaks; consider relaxing the rule to reduce false positives.

---

Learnt from: github-actions[bot]
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:58:09.061Z
Learning: In OmniBlocks/scratch-gui, when cleaning PR threads, prefer minimizing CodeRabbit auto-status comments that contain phrases like “For best results, initiate chat…”, “Rate Limit Exceeded”, and “This is an automated status update…”, while leaving human-directed replies intact.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-12-05T16:55:31.105Z
Learning: OmniBlocks/scratch-gui: supervoidcoder wants automated comment moderation to detect spam/phishing from first-time commenters using GitHub Models AI (similar to ulti.yml workflow). Should analyze comments and hide/delete based on severity levels.

---

Learnt from: github-actions[bot]
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:58:09.061Z
Learning: Extend spam-cleanup patterns in OmniBlocks/scratch-gui to include “auto-generated comment: … by coderabbit.ai” and structural markers (tips_start, analysis chain, agentComponentType, rawResChunk) to fully minimize CodeRabbit auto-noise.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:55:33.086Z
Learning: OmniBlocks/scratch-gui: Never reply to CodeRabbit’s own auto-generated comments (status, tips, rate-limit); when a bot loop is detected on a PR, post a single apology, minimize CodeRabbit auto-noise, and then remain silent unless a human asks for more.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:34:24.309Z
Learning: CodeRabbit experienced a spam bug on PR `#309` in OmniBlocks/scratch-gui where it posted 11 duplicate "no user input to reply to" messages. The spam was triggered by bot comments from github-actions and wellcode-ai. The comments were successfully minimized using GitHub GraphQL API mutations with the OFF_TOPIC classifier.

---

Learnt from: github-actions[bot]
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:50:28.065Z
Learning: OmniBlocks/scratch-gui: When cleaning PR noise, prefer minimizing CodeRabbit auto-generated comments matching headers like "This is an auto-generated reply by CodeRabbit", "Rate Limit Exceeded", and "automated status update", while preserving human-directed conversation.

---

Learnt from: 8to16
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2026-01-22T19:24:58.994Z
Learning: OmniBlocks/scratch-gui moderation preference: When CodeRabbit detects a human comment with strong profanity or a clear Code of Conduct violation (e.g., slurs, harassment, threats, sexual content, doxxing), it must ping 8to16 and supervoidcoder in the reply. Do not ping for mild words like "crap" or "heck" (explicitly allowed by project policy). Do not trigger on bot/CI comments. Keep the tone calm and optionally include a brief CoC reminder. This rule applies only within the OmniBlocks/scratch-gui repository.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2026-01-27T22:15:17.812Z
Learning: In OmniBlocks/scratch-gui, CodeRabbit should only respond to comments that directly address coderabbitai. If a comment is directed at another bot (e.g., copilot) but mentions CodeRabbit in passing, CodeRabbit should not respond - the comment is meant for the other bot to handle.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2026-01-02T20:01:55.528Z
Learning: In OmniBlocks/scratch-gui, when a thread gets noisy, supervoidcoder prefers deleting CodeRabbit auto-generated plan/log comments on that thread (issue/PR) while keeping human/chatty replies; for Amazon Q boilerplate, prefer minimizing rather than deleting unless told otherwise.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-10-10T15:23:20.794Z
Learning: OmniBlocks/scratch-gui PR test workflow now ignores bot/AI comments containing the trigger phrase; only human users are allowed to trigger the deployment.

---

Learnt from: supervoidcoder
Repo: OmniBlocks/scratch-gui PR: 0
File: :0-0
Timestamp: 2025-11-05T01:48:23.435Z
Learning: OmniBlocks/scratch-gui: For CI comment hygiene, prefer not mentioning bot/app requesters in the Mega PR Test & Analysis sticky comment; only mention humans. Track such behavior changes via a GitHub issue assigned to supervoidcoder with backlinks to the initiating PR and comment.

[supervoidcoder]

@ampelectrecuted Whre did this come from? Like why did you make this
Was there another flanker that came on our repo to spAm ?

@coderabbitai