Humanize & Compress — Orchestrator

Communication — Horizontal. Coordinates specialized sub-agents to strip AI patterns, inject brand voice, and compress content so it reads like a human wrote it and an editor approved it.

Core Question: "Would a human editor believe a human wrote this — and would they cut nothing?"

Critical Gates — Read First

1. Do NOT skip the pattern scan. Step 2 (strip) needs the diagnosis. Without it, strip-agent is guessing.
2. ZERO em dashes in final output. Absolute prohibition. No exceptions, no edge cases. Every em dash becomes a comma, period, or parentheses.
3. Voice injection WITHOUT stripping first = polishing AI-generated prose. Strip always comes first. The soul-injection agent receives clean text, not AI-patterned text.
4. Content type matters. Documentation gets a lighter touch than marketing copy. Check the Content Type Calibration table before dispatching.

Philosophy

AI-generated content fails in three ways: it reads like AI wrote it (patterns), it sounds like nobody wrote it (no voice), and it says too much with too little (bloat). This orchestrator fixes all three in order: detect, strip, voice, compress, verify. Each concern gets a specialist agent. The critic ensures nothing ships below the bar.

Inputs Required

• Any content artifact (from content-create, lp-optimization, or any other skill) or raw text