Building an Auto-Generated Technical Blog System Across Four Domains

```html

This session focused on creating a granular, session-aware technical documentation system that automatically generates blog posts at tech.queenofsandiego.com, tech.dangerouscentaur.com, tech.sailjada.com, and tech.burialsatseasandiego.com. The goal: provide complete visibility into all technical work performed across these properties, with enough detail to satisfy rigorous technical review.

Architecture Overview

The system captures session transcripts from Claude Code's project directories, extracts work performed, redacts sensitive data, and generates detailed HTML blog posts. Each post is published to its respective domain via S3 + CloudFront infrastructure.

Session Capture: Claude Code stores JSONL-formatted session transcripts in ~/.claude/projects with automatic file write/edit tracking
Post Generation: Python scripts parse transcripts, extract tool usage and file modifications, generate HTML
Publishing: Posts uploaded to domain-specific S3 buckets, CloudFront invalidations trigger cache refresh
Integration: Ship's Papers navigation menus updated to link to tech blogs

Infrastructure Setup

Four parallel deployments were created using existing wildcard SSL certificates and domain infrastructure:

queenofsandiego.com: Wildcard cert already present; created tech-qos S3 bucket and CloudFront distribution
sailjada.com: Wildcard cert already present; created tech-jada S3 bucket and CloudFront distribution with Route53 alias records
dangerouscentaur.com: Wildcard CloudFront distribution E2Q4UU71SRNTMB already serving dc-sites bucket; added tech.* origin path routing
burialsatseasandiego.com: GoDaddy-managed DNS; created tech-bats S3 bucket, CloudFront distribution, and added CNAME record to GoDaddy DNS via API

All distributions configured with:

HTTPS default, HTTP redirects to HTTPS
Index document: index.html
Error document: index.html (for client-side routing)
Gzip compression enabled for text/HTML assets

Session Transcript Processing

Claude Code stores session data in ~/.claude/projects/-Users-cb-Documents-repos/sessions as JSONL files. Each session contains:

{
  "type": "tool_use",
  "name": "write",
  "input": {"path": "/path/to/file"},
  "output": "File written successfully"
}

The blog generator script (tech_blog_generator.py) extracts:

All write and edit operations with full file paths
Shell commands executed with their arguments (secrets redacted via regex patterns)
File reads that provide context
Tool outputs indicating success/failure

A sensitive data redaction layer removes:

AWS Access Key patterns
API tokens and bearer credentials
Email addresses and personal identifiers
Database passwords and connection strings
ACM certificate ARNs (replaced with [REDACTED_CERT_ARN])

Blog Post Generation Logic

Generated posts follow a consistent structure:

Title: Extracted from session objective or file operations pattern
Files Modified: Detailed list with relative paths, operation type (create/edit), and brief context
Commands Executed: AWS CLI, Python script invocations, shell operations with arguments visible (credentials removed)
Infrastructure Changes: S3 buckets created, CloudFront distributions, Route53 records, ACM certificates validated
Technical Decisions: Why certain tools/patterns were chosen
Validation: Test commands run, expected outputs

Each post is timestamped and indexed in an index.json file at the blog root for easy navigation and programmatic access.

Stop Hook Integration

A critical component: the Claude Code stop hook (/Users/cb/.claude/hooks/tech_blog_stop.sh) triggers at session end. This script:

Reads the most recent session transcript from ~/.claude/projects/-Users-cb-Documents-repos/sessions
Invokes the Python blog generator with the transcript path
Determines target domain based on session context (repo paths, memory files referenced)
Uploads generated HTML to the appropriate S3 bucket
Invalidates CloudFront distribution cache
Logs all operations to ~/.claude/logs/tech_blog.log

The hook is registered in ~/.claude/settings.json under the project configuration, ensuring it runs for all sessions in the repos project.

Navigation Integration

The Ship's Papers menu structure was updated in queenofsandiego.com/index.html to include a "Technical Documentation" submenu item linking to https://tech.queenofsandiego.com. Similar updates made to dangerouscentaur, sailjada, and burialsatseasandiego navigation menus.

Testing & Validation

Initial post generated from this session's transcript. Key validations:

File paths in post match actual modifications (verified against write operations)
Sensitive data properly redacted (credentials, API keys, certificate ARNs removed)
S3 uploads successful; CloudFront distributions reporting healthy status
DNS propagation complete (verified via nslookup for tech.queenofsandiego.com, etc.)
HTTP/HTTPS access working with proper redirects

Known Issues & Follow-up Work

Image Assets: burialsatseasandiego.sailjada.com displaying incorrect images for "imagine" and "small catamaran" — task created on progress board for fleet image replacement
Guest Count Messaging: Minimum 10 guests text across booking pages needs consistency review and update
GA4 Audit: Identified multiple GA tracking IDs across properties; comprehensive audit and booking optimization recommendations pending separate analysis

What's Next

The system is now live and automated. Each development session will automatically generate a granular technical blog post capturing:

Exact files touched and operations performed
Infrastructure resources created/modified with full resource identifiers
Commands executed with visible arguments (no secrets)
Architecture