Building Multi-Site Automated Technical Blog Infrastructure with Session-Based Post Generation

```html

What Was Done

Implemented a complete automated technical blogging system across four separate domains (queenofsandiego.com, sailjada.com, dangerouscentaur.com, and burialsatseasandiego.com) that automatically captures development work from Claude Code sessions and publishes granular technical posts without exposing credentials or sensitive data.

Architecture Overview

The system consists of three core components:

Session Capture Hook: A Stop hook integrated into Claude Code that runs when development sessions end
Infrastructure Init Script: Provisions S3 buckets, CloudFront distributions, and DNS records for each tech blog domain
Blog Generator: Parses session transcripts, extracts technical work, and generates HTML posts with automatic credential filtering

Infrastructure Setup

S3 Bucket Architecture

Created dedicated S3 buckets for each tech blog:

qos-tech-blog — Hosts tech.queenofsandiego.com
jada-tech-blog — Hosts tech.sailjada.com
dc-sites — Existing bucket repurposed with wildcard CloudFront distribution for *.dangerouscentaur.com
bats-tech-blog — Hosts tech.burialsatseasandiego.com

Each bucket is configured as a static website host with index.html as the default root object. Buckets are blocked from public ACL access; all traffic routes through CloudFront.

CloudFront Distribution Strategy

Four CloudFront distributions provide global edge caching:

Three new distributions point to qos-tech-blog, jada-tech-blog, and bats-tech-blog S3 origins
Existing dangerouscentaur distribution already provides wildcard coverage for all *.dangerouscentaur.com subdomains
All distributions use AWS Certificate Manager wildcard certificates (*.queenofsandiego.com, *.sailjada.com, and wildcard for dangerouscentaur)
Cache behavior set to 300-second TTL for HTML to balance freshness with performance

DNS Configuration

Different DNS providers required different approaches:

queenofsandiego.com (Route53): Created tech.queenofsandiego.com CNAME record pointing to CloudFront distribution domain
sailjada.com (Route53): Created tech.sailjada.com CNAME record pointing to CloudFront distribution domain
dangerouscentaur.com (Namecheap): Added CNAME for tech.dangerouscentaur.com pointing to existing wildcard distribution
burialsatseasandiego.com (GoDaddy): Added CNAME for tech.burialsatseasandiego.com pointing to CloudFront; also integrated ACM certificate DNS validation records

Session Hook Integration

The Stop hook is located at /Users/cb/.claude/hooks/tech_blog_stop.sh and runs automatically when Claude Code sessions terminate. The hook:

Exports the current session transcript in JSONL format from Claude Code's internal storage
Passes the transcript to the blog generator with site context (determining which domain to publish to based on file modifications)
Filters out any sensitive patterns (API keys, tokens, credentials, personal identifiers)
Generates a timestamped HTML post and uploads it to the appropriate S3 bucket via AWS CLI
Invalidates the CloudFront distribution cache to ensure new posts appear immediately
Logs all operations to ~/.claude/logs/tech_blog_generation.log for audit trails

Blog Generator Logic

The generator (/Users/cb/Documents/repos/tools/tech_blog_generator.py) parses session transcripts by:

Identifying modified files: Examines the "Files modified/created" section to determine which site(s) were worked on
Extracting tool invocations: Parses JSONL records where type: "tool_use" to capture commands run (with credential filtering)
Building semantic context: Groups related file modifications and commands into logical work units (e.g., "Infrastructure Setup" vs "Content Updates")
Creating HTML structure: Generates semantic HTML with <h2>, <h3>, and <code> tags for readability
Sanitizing output: Strips passwords, API keys, tokens, secret ARNs, and personal data using regex patterns and keyword matching
Timestamping: Uses ISO 8601 timestamps for post publication dates and file organization

Navigation Integration

Updated /Users/cb/Documents/repos/sites/queenofsandiego.com/index.html to add a "Ship's Papers" menu section with a "Technical Blog" link pointing to tech.queenofsandiego.com. This provides user-visible access to development transparency. Similar navigation will be added to other main sites.

Key Technical Decisions

Why Separate S3 Buckets Instead of One Shared Bucket

Separate buckets provide clear ownership boundaries, independent scaling, and easier access control policies. If one site's traffic spikes, it doesn't affect others. Disaster recovery is also isolated—rolling back one blog doesn't require understanding another site's history.

Why CloudFront Over Direct S3 Access

CloudFront provides geographic distribution (faster loads for global users), DDoS protection via AWS Shield Standard, and the ability to add request/response headers without touching S3. HTTPS termination at the edge is also essential for security.

Why Session-Based Rather Than Git-Hook-Based

Claude Code sessions are self-contained and preserve full context (exact prompts, tool outputs, file contents). Git hooks capture only commit messages and diffs. Sessions capture the reasoning and decision-making process, which is valuable for Sergio to understand not just what changed, but why.

Credential Filtering Strategy

The generator uses multiple filtering layers: regex patterns for AWS ARNs, API Gateway keys, and JWT tokens; keyword matching for "password," "secret," "key," and "token"; and redaction of sensitive environment variables. This defense-in-depth approach minimizes the risk of accidental exposure.

Infrastructure Code

The initialization script /Users/cb/Documents/repos/tools/tech_blog_init.py is idempotent and handles:

S3 bucket creation with versioning and website hosting configuration
CloudFront distribution creation with appropriate cache behaviors
DNS