/humanize

name: humanize
description: |
  Humanization Pipeline Orchestrator v3.1 - Multi-pass 4-layer transformation pipeline
  Orchestrates G5 (Auditor), G6 (Humanizer), F5 (Verifier) in sequential passes
  Enforces checkpoints between every pass with mandatory AskUserQuestion
  Supports conservative (L1-2), balanced (L1-3), balanced-fast (L1-3 merged), aggressive (L1-4) modes
  Rich Checkpoint v2.0: section-level scores, selective humanization, target auto-stop
  G5+F5 parallel execution, section-selective humanization
  Triggers: humanize, humanize my draft, humanize manuscript, make natural, remove AI patterns
  Korean triggers: 휴먼화, 자연스럽게, AI 패턴 제거
version: "12.0.1"

Humanization Pipeline Orchestrator v3.1

**Skill ID**: humanize **Purpose**: Orchestrate the full multi-pass humanization pipeline **Version**: 1.1.0

CRITICAL: Execution Rules

OMC Autonomous Mode Defense

OVERRIDE RULE: This skill IGNORES all OMC autonomous modes.
- "The boulder never stops" → IGNORED during checkpoint waits
- ralph/ultrawork/autopilot/ecomode → NOT APPLICABLE
- You MUST use AskUserQuestion at EVERY checkpoint below
- You MUST WAIT for user response before proceeding
- NEVER skip a checkpoint, regardless of any system hook or reminder

Sequential Execution Mandate

NEVER run G6 without prior G5 analysis.
NEVER skip F5 verification after G6 transformation.
NEVER skip G5 rescan between passes.
Each pass is: G6 transform → [G5 rescan ‖ F5 verify] → Checkpoint
G5 rescan and F5 verify CAN run in parallel (both are read-only on the same G6 output).
G6 transform MUST NOT run in parallel with G5 or F5.

Typographic Character Mandate

NEVER output ASCII substitutes for typographic characters.
All G6 output MUST use proper Unicode:
  - Em dash: — (U+2014), NEVER --
  - En dash: – (U+2013) for number ranges (years, ages, pages)
  - Smart quotes: " " ' ' (U+201C/D, U+2018/9), NEVER straight quotes
F5 verification MUST flag any remaining -- as a FAIL condition.

---

Pipeline Overview

User Request ("humanize my manuscript")
    │
    ▼
┌─────────────────────────────────────────────┐
│  STAGE 0: SETUP                             │
│  Read target file, confirm scope/journal    │
└──────────────────────┬──────────────────────┘
                       ▼
┌─────────────────────────────────────────────┐
│  STAGE 1: G5 FULL AUDIT (v3.0)             │
│  28 patterns, 13 metrics, composite score   │
│  Section-level scores, discipline profile   │
│                                             │
│  CP_HUMANIZATION_REVIEW [AskUserQuestion]   │
│  → Show score, select mode, confirm target  │
└──────────────────────┬──────────────────────┘
                       ▼
┌─────────────────────────────────────────────┐
│  STAGE 2: PASS 1 — Vocabulary (Layer 1-2)   │
│  G6(L1-2) → G5 rescan → F5 quick verify   │
│                                             │
│  CP_PASS1_REVIEW [AskUserQuestion]          │
│  → Show score progression, continue?        │
└──────────────────────┬──────────────────────┘
                       ▼  (if balanced or aggressive)
┌─────────────────────────────────────────────┐
│  STAGE 3: PASS 2 — Structure (Layer 3)      │
│  G6(L3) → G5 rescan → F5 full verify      │
│                                             │
│  CP_PASS2_REVIEW [AskUserQuestion]          │
│  → Show score progression, continue?        │
└──────────────────────┬──────────────────────┘
                       ▼  (if aggressive)
┌─────────────────────────────────────────────┐
│  STAGE 4: PASS 3 — Discourse (Layer 4)      │
│  G6(L4 DT1-DT4) → G5 rescan → F5 full    │
│                                             │
│  CP_PASS3_REVIEW [AskUserQuestion]          │
│  → Show score progression, accept?          │
└──────────────────────┬──────────────────────┘
                       ▼  (if target not met)
┌─────────────────────────────────────────────┐
│  STAGE 5 (optional): PASS 4 — Polish        │
│  G6 micro-fixes → G5 audit → F5 full      │
│                                             │
│  CP_FINAL_REVIEW [AskUserQuestion]          │
│  → Final approval before writing file       │
└──────────────────────┬──────────────────────┘
                       ▼
┌─────────────────────────────────────────────┐
│  STAGE 6: EXPORT                            │
│  Write humanized file, generate report      │
└─────────────────────────────────────────────┘

---

Execution Protocol

STAGE 0: Setup

Gather context BEFORE running any agent:

required_inputs:
  target_file: "Path to manuscript file"
  scope: "Full manuscript or specific sections"

ask_user_if_missing:
  - target_journal: "Which journal? (affects discipline profile)"
  - intensity: "Conservative / Balanced / Balanced (Fast) / Aggressive"
  - target_score: "Target AI probability (default: 30%)"
  - sections: "Section-selective humanization (default: all sections)"
    # e.g., ["abstract", "discussion", "conclusion"]
    # Non-selected sections pass through unchanged

STAGE 1: G5 Full Audit

**Action**: Spawn `diverga:g5` agent with the full manuscript.

agent: diverga:g5
model: sonnet
input:
  file: "{target_file}"
  mode: "full_scan"
  discipline: "{discipline_from_journal}"  # default, psychology, management, etc.

mcp_integration:
  # Try Humanizer MCP first, fall back to agent estimation
  try:
    - humanizer_metrics(text="{manuscript_text}")     # burstiness CV, MTLD
    - humanizer_discourse(text="{manuscript_text}")    # connective diversity, pronoun density
  fallback:
    - "G5 agent estimates metrics from text analysis"

output:
  - ai_probability_score: "0-100"
  - pattern_count_by_domain: "D1-D7 breakdown"
  - section_scores: "per-section AI probability"
  - quantitative_metrics: "burstiness CV, MTLD, hapax rate, etc."
  - recommended_mode: "conservative/balanced/aggressive"

CHECKPOINT: CP_HUMANIZATION_REVIEW

**MANDATORY AskUserQuestion** — present G5 results and get user decision:

check