Automatically publish GitHub commits as blog posts

What This Agent Does

This powerful automation agent transforms your blog content creation process by automatically generating SEO-optimized WordPress posts complete with AI-generated featured images. When triggered via webhook, it uses Claude AI to craft compelling product-focused content, generates a custom image prompt, creates a stunning visual using Google's Imagen 4, and publishes everything directly to your WordPress site—all without manual intervention.

Key benefits: Save 2-3 hours per blog post by eliminating manual writing, image sourcing, and formatting. Maintain consistent brand voice and quality across all content. Scale your content production effortlessly while ensuring every post includes professional, contextually relevant imagery.

Perfect for: E-commerce businesses launching new products, content marketing teams managing multiple blogs, digital agencies serving clients, and entrepreneurs building their online presence through consistent, high-quality content publication.

Required Integrations

Anthropic (Claude AI)

Why it's needed: Claude AI powers both the content generation and image prompt creation phases of this workflow. It writes your blog post content with expert product copywriting skills and crafts detailed, effective prompts for the image generation system.

Setup steps:

1. Visit console.anthropic.com and create an account or sign in
2. Navigate to API Keys in the left sidebar
3. Click Create Key and give it a descriptive name like "TaskAGI Blog Automation"
4. Copy the API key immediately (it won't be shown again)
5. Store it securely in a password manager

Configuration in TaskAGI:

1. Go to Integrations in your TaskAGI dashboard
2. Search for "Anthropic" and click Connect
3. Paste your API key in the API Key field
4. Click Test Connection to verify
5. Save the integration with a recognizable name like "Anthropic - Blog Content"

Important: Anthropic charges per token usage. The Claude Haiku model used in this workflow is cost-effective, typically costing $0.01-0.05 per blog post depending on length.

Google Imagen 4 (Fast)

Why it's needed: Google's Imagen 4 generates high-quality, photorealistic featured images based on the AI-crafted prompts. The "Fast" variant provides excellent quality with reduced generation time, perfect for blog workflows.

Setup steps:

1. Go to console.cloud.google.com
2. Create a new project or select an existing one
3. Enable the Vertex AI API from the API Library
4. Navigate to IAM & Admin → Service Accounts
5. Click Create Service Account
6. Name it "TaskAGI Image Generation" and grant it the Vertex AI User role
7. Create and download a JSON key file
8. Enable billing for your project (Imagen 4 requires an active billing account)

Configuration in TaskAGI:

1. Navigate to Integrations → Add Integration
2. Select Google Imagen 4 Fast
3. Upload your service account JSON key file
4. Specify your Project ID (found in Google Cloud Console)
5. Set your preferred Region (us-central1 recommended for best availability)
6. Test the connection and save

Pricing note: Google Imagen 4 Fast costs approximately $0.02-0.04 per image generated. Monitor your usage through the Google Cloud Console.

WordPress

Why it's needed: This integration publishes your completed blog posts directly to your WordPress site, including the generated content, featured image, and all formatting—eliminating manual copy-paste work.

Setup steps:

1. Log into your WordPress admin dashboard
2. Install the Application Passwords plugin (if using WordPress 5.6+, this is built-in)
3. Go to Users → Profile
4. Scroll to Application Passwords section
5. Enter "TaskAGI Automation" as the application name
6. Click Add New Application Password
7. Copy the generated password immediately (spaces can be included or removed)
8. Note your WordPress site URL and username

Configuration in TaskAGI:

1. Open Integrations and select WordPress
2. Enter your Site URL (e.g., https://yourblog.com)
3. Input your Username
4. Paste the Application Password you generated
5. Click Test Connection to verify access
6. Save with a descriptive name like "WordPress - Main Blog"

Security tip: Application passwords can be revoked anytime from your WordPress profile without affecting your main login credentials.

Configuration Steps

Step 1: Configure the Webhook Trigger

The webhook trigger initiates your workflow whenever it receives a POST request with product information.

Configuration:

1. Click on the Webhook node (first node in your workflow)
2. Copy the Webhook URL provided (it looks like https://api.taskagi.com/webhook/abc123...)
3. Set Authentication to your preferred security level:
   • None: Open webhook (use only for testing)
   • API Key: Requires a key in the request header (recommended)
   • HMAC Signature: Most secure option for production

Expected payload structure:

{
  "product_name": "Wireless Noise-Canceling Headphones",
  "product_category": "Electronics",
  "key_features": ["40-hour battery", "Active noise cancellation", "Bluetooth 5.0"],
  "target_audience": "Remote workers and commuters",
  "price_point": "$199"
}

Testing tip: Use a tool like Postman or curl to send test payloads during setup.

Step 2: Set Up the Conditional Check (If Condition Node)

This node validates that your webhook received the necessary product information before proceeding.

Configuration:

1. Select the If Condition node
2. Set the condition to check for required fields:
   • Field: {{trigger.body.product_name}}
   • Operator: is not empty
3. Add additional conditions using AND logic:
   • {{trigger.body.product_category}} is not empty
   • {{trigger.body.key_features}} is not empty

Why this matters: This prevents the workflow from wasting API credits on incomplete data and ensures quality output.

Step 3: Configure Content Generation (First Claude Node)

This node creates your blog post content using Claude Haiku's expert writing capabilities.

Configuration:

1. Click the Generate Text (Claude) node
2. Verify the Model is set to claude-haiku-4-5 (fast and cost-effective)
3. Review the system prompt (already configured):

You are an expert product content writer for TaskAGI...

4. Customize the User Message to include webhook data:

Write a comprehensive blog post about {{trigger.body.product_name}}.

Product Category: {{trigger.body.product_category}}
Key Features: {{trigger.body.key_features}}
Target Audience: {{trigger.body.target_audience}}
Price Point: {{trigger.body.price_point}}

Create an engaging, SEO-optimized post with:
- Compelling headline
- Introduction hook
- Feature breakdown
- Benefits for target audience
- Call-to-action conclusion

Length: 800-1200 words
Tone: Professional yet conversational

5. Set Max Tokens to 4096 (allows for longer content)
6. Set Temperature to 0.7 (balanced creativity and consistency)

Output: This node produces {{nodes.2957.content}} containing your complete blog post text.

Step 4: Configure Image Prompt Generation (Second Claude Node)

This specialized node creates detailed, effective prompts for the image generator.

Configuration:

1. Select the Prompt for image generation node
2. Keep the model as claude-haiku-4-5
3. The system prompt is pre-configured with image generation best practices:

# AI Image Generation Prompt for Blog Posts

## Instructions:
Create a detailed, specific image generation prompt based on the blog content...

4. Set the User Message to:

Based on this blog post content, create an image generation prompt:

{{nodes.2957.content}}

Product: {{trigger.body.product_name}}
Category: {{trigger.body.product_category}}

5. Set Max Tokens to 500 (sufficient for detailed prompts)
6. Set Temperature to 0.8 (more creative for visual descriptions)

Output: Produces {{nodes.2958.content}} with an optimized image prompt like: "Professional product photography of wireless headphones on a minimalist desk setup, soft natural lighting, modern workspace aesthetic, high detail, 4K quality"

Step 5: Configure Image Generation (Google Imagen Node)

This node transforms the AI-crafted prompt into a stunning featured image.

Configuration:

1. Click the Generate Image (Fast) node
2. Set Prompt to: {{nodes.2958.content}}
3. Configure image parameters:
   • Aspect Ratio: 16:9 (ideal for blog featured images)
   • Number of Images: 1
   • Safety Filter: medium (blocks inappropriate content)
4. Advanced settings (optional):
   • Guidance Scale: 7.5 (balance between prompt adherence and creativity)
   • Negative Prompt: "blurry, low quality, watermark, text overlay"

Output: Generates {{nodes.2959.image_data}} containing the base64-encoded image data.

Step 6: Configure WordPress Post Creation

The final node publishes everything to your WordPress site automatically.

Configuration:

1. Select the Create WordPress Post node
2. Configure post details:
   • Title: {{nodes.2957.content | extractTitle}} or manually set from webhook data
   • Content: {{nodes.2957.content}}
   • Status: Choose from:
     • publish (immediately live)
     • draft (requires manual review)
     • pending (awaits editor approval)
3. Set Featured Image:
   • Image Source: From URL or Base64
   • Image Data: {{nodes.2959.image_data}}
4. Configure categories and tags:
   • Categories: {{trigger.body.product_category}}
   • Tags: Extract from content or set manually
5. SEO settings (if using Yoast/RankMath):
   • Meta Description: First 155 characters of content
   • Focus Keyword: {{trigger.body.product_name}}

Pro tip: Start with draft status while testing, then switch to publish for production use.

Step 7: Configure the No Operation Node

This node handles the "false" path when the conditional check fails.

Configuration:

1. Click the No Operation node
2. Optionally add logging:
   • Enable Log Message: "Webhook received incomplete data"
   • Set Log Level: warning
3. Consider adding a notification integration here to alert you of failed validations

Testing Your Agent

Running Your First Test

1. Prepare test data: Create a JSON payload with sample product information:

{
  "product_name": "EcoSmart Water Bottle",
  "product_category": "Sustainable Living",
  "key_features": ["BPA-free", "Keeps drinks cold 24hrs", "Recycled materials"],
  "target_audience": "Environmentally conscious consumers",
  "price_point": "$29.99"
}

2. Send the webhook request:

   • Use Postman, Insomnia, or curl
   • POST to your webhook URL
   • Include your authentication if configured
   • Set Content-Type header to application/json

3. Monitor execution:

   • Open the Workflow Runs tab in TaskAGI
   • Watch the real-time execution progress
   • Each node will show green checkmarks as it completes

Verification Checklist

After the If Condition node:

• ✅ Verify the condition evaluated to true
• ✅ Check that all required fields were present
• ✅ Confirm the workflow proceeded to content generation

After content generation:

• ✅ Review the generated blog post in the node output
• ✅ Verify it includes all sections (intro, features, benefits, CTA)
• ✅ Check word count is within 800-1200 range
• ✅ Ensure product name and features are accurately incorporated

After image prompt generation:

• ✅ Read the generated prompt for clarity and detail
• ✅ Confirm it references the product appropriately
• ✅ Verify it includes style and quality descriptors

After image generation:

• ✅ Preview the generated image in the node output
• ✅ Check that it's relevant to the product and content
• ✅ Verify image quality and resolution are acceptable
• ✅ Ensure no inappropriate or off-brand elements appear

After WordPress post creation:

• ✅ Log into your WordPress dashboard
• ✅ Navigate to Posts → All Posts
• ✅ Open the newly created post
• ✅ Verify the content formatted correctly
• ✅ Confirm the featured image uploaded successfully
• ✅ Check categories and tags applied properly
• ✅ Preview the post as it will appear to visitors

Expected Results

Successful execution should:

• Complete in 30-60 seconds total
• Generate 800-1200 words of coherent, on-brand content
• Produce a relevant, high-quality featured image
• Create a properly formatted WordPress post
• Show all green status indicators in the workflow run log

Performance benchmarks:

• Content generation: 10-15 seconds
• Image prompt creation: 3-5 seconds
• Image generation: 15-25 seconds
• WordPress publishing: 5-10 seconds

Troubleshooting

Webhook Not Triggering

Symptoms: No workflow execution appears after sending webhook request

Solutions:

• Verify the webhook URL is correct (copy it again from the node)
• Check authentication credentials if using API key or HMAC
• Ensure Content-Type header is set to application/json
• Verify your payload is valid JSON (use a JSON validator)
• Check TaskAGI's webhook logs for rejected requests

Test command:

curl -X POST https://your-webhook-url \
  -H "Content-Type: application/json" \
  -d '{"product_name":"Test Product","product_category":"Test"}'

If Condition Always Goes to False Path

Symptoms: Workflow stops at the No Operation node every time

Solutions:

• Inspect the webhook payload in the trigger node output
• Verify field names match exactly (case-sensitive)
• Check for extra spaces or special characters in field names
• Use the expression tester: {{trigger.body}} to see all received data
• Ensure you're sending data in the request body, not query parameters

Debug tip: Temporarily change the condition to always true to test downstream nodes.

Content Generation Fails or Returns Errors

Symptoms: Claude node shows red error status, or content is truncated

Common errors and fixes:

"Authentication failed"

• Regenerate your Anthropic API key
• Verify no extra spaces when pasting the key
• Check your Anthropic account has available credits

"Rate limit exceeded"

• You've hit Anthropic's rate limits
• Wait 60 seconds and retry
• Consider upgrading your Anthropic plan
• Add a delay node before this step if running multiple workflows

"Max tokens exceeded"

• Reduce the Max Tokens setting to 3000
• Simplify your prompt to require less output
• Break content generation into multiple steps

Content quality issues:

• Adjust temperature (lower = more focused, higher = more creative)
• Refine your system prompt with more specific instructions
• Add examples of desired output format in the prompt
• Include more context from the webhook data

Image Generation Produces Irrelevant Images

Symptoms: Generated images don't match the product or content

Solutions:

• Review the image prompt in node 2958's output
• Refine the prompt generation instructions to be more specific
• Add negative prompts to exclude unwanted elements
• Increase guidance scale to 10-12 for stricter prompt adherence
• Include more product-specific details in the prompt template

Prompt improvement tips:

• Add specific style descriptors: "product photography style," "lifestyle shot," "studio lighting"
• Include color schemes: "warm tones," "minimalist