Ghost CMS Demo Deployment – Ghost in the Machine

This repository contains a customized Ghost theme and deployment setup used to demonstrate CMS workflows. It also implements a content automation pipeline using n8n, OpenAI, and GitHub.

Purpose

This project demonstrates:

• Rapid configuration and customization of a Ghost CMS theme
• Version-controlled deployment to a live Render instance
• Integration of lightweight analytics and content-driven navigation
• A developer-friendly publishing pipeline
• Fluent orchestration of CMS, automation, and AI systems

Architecture

• Ghost CMS deployed via Render
• GitHub – source of truth for theme files and routing config
• n8n (cloud-hosted) – orchestration layer for inbound content and logic
• OpenAI – summarization and contextual transformation layer (in development)
• Plausible Analytics – integrated for privacy-respecting metrics
• Serper.dev – reserved for future use in search-aware prompting and source enrichment

Theme

The theme is a lightly modified fork of Ghost’s Dawn theme, with changes focused on:

• External link handling and improved metadata
• Analytics injection (Plausible, with domain-specific logic)
• Routing and homepage customization via routes.yaml
• Git-backed development via entrypoint.sh

Deployment Workflow

• Theme and routing files are versioned in this repository
• Pushes to the main branch trigger Render builds when theme or configuration files are updated
• Render applies the updated theme automatically during deployment
• Environment variables configure the Ghost instance dynamically

=======

Required Environment Variables

The container expects the following variables at runtime:

• PORT – network port exposed by Ghost
• MAIL__SMTP__AUTH__PASS – SMTP password for outbound mail
• URL – canonical site URL used in config.production.json

Live Instance

https://adambalm.io

Persona Activation and Development Constraints

This repository operates under persistent protocols to ensure clarity, cost control, and architectural discipline.

• ScrumMaster persona is activated via scrum, scrummaster, or close variants. It enforces all project constraints and provides ongoing guidance in agile methodology.
• Black Flag protocol is activated explicitly or automatically when output risks speculative reasoning or satisfaction bias. It prioritizes epistemological hygiene.
• Red Team mode simulates adversarial testing of ideas, systems, and architecture to expose weak points.

Operating principles include:

• Mise-en-place: Tasks begin only after confirming all credentials, tokens, and access are in place.
• Cost awareness: Active tracking of Render, OpenAI, GitHub Actions, and n8n usage. No polling, triggers, or API calls that risk overage.
• No PII: No personally identifiable information is ever to be stored, committed, or displayed. This constraint applies across all personae unless explicitly overridden.

Current Status

• Render-hosted Ghost CMS is live and thematically customized
• Webhook-based n8n flow has been verified via manual POST
• Webhook remains unauthenticated
• OpenAI logic not yet connected
• GitHub Action for scheduled RSS fetch (to replace blocked direct fetch) not yet implemented

Next Steps

• Configure authentication for n8n webhook
• Wire OpenAI for summarization and persona-based analysis
• Create a GitHub Action to fetch Substack posts and POST to n8n
• Add citation handling and excerpt logic for source attribution
• Initiate critical-project-prompts.md as a canonical prompt reference for this and future repos
• Expand scope of integration with Serper.dev for search context and source visibility enhancements
• Change ridiculous author photo

Epistemelogical Controls

The underlying cognitive limits and system assumptions imposed on LLMs assisting this project:

black-flag.md

License

MIT (inherited from original theme)