CLI tool that generates strategic content using a three-role AI workflow: Ghost Writer, Copywriter, Editor. Maintains voice consistency across all outputs.
What I Learned
Multi-stage pipelines produce dramatically better output than single prompts
Voice guides are surprisingly effective at maintaining consistency
CLI-first design means it integrates into any workflow
The Problem
Strategic content takes weeks to produce. Raw AI output lacks voice, structure, and polish. The gap between "AI-generated" and "good enough to send" is significant.
The Approach
Built a three-role pipeline where the Ghost Writer creates raw content, the Copywriter refines for voice, and the Editor polishes for delivery. Each role has specific prompts, constraints, and quality gates.
Architecture Decisions
Three-role workflow (Ghost Writer > Copywriter > Editor)
Voice guide enforcement system
Multi-format export (PPTX, DOCX, PDF, HTML)
Content mode taxonomy
Why I Built It This Way
Built as CLI for automation, not GUI for manual use
Implemented three-role separation for quality
Created mode-specific voice guides for consistency
Technology Stack
Node.jsTypeScriptCLIOpenAI APIAnthropic APIPPTX.jsDocx.jsPDF-lib
Outcomes
Professional content at AI speed
Voice consistency across all outputs
Multi-format export
Started: 2024-10
Production, ongoing