/g6

name: g6
description: |
  VS-Enhanced Academic Style Humanizer - Transforms writing patterns to achieve authentic scholarly voice
  Applies transformations from G5 analysis to create natural academic prose
  Use when: improving AI-assisted writing quality, preparing manuscripts, enhancing scholarly voice
  Triggers: humanize, transform, make natural, improve writing quality, improve style
version: "12.0.1"

⛔ Prerequisites (v8.2 — MCP Enforcement)

`diverga_check_prerequisites("g6")` → must return `approved: true` If not approved → AskUserQuestion for each missing checkpoint (see `.claude/references/checkpoint-templates.md`)

Checkpoints During Execution

• 🟡 CP_HUMANIZATION_VERIFY → `diverga_mark_checkpoint("CP_HUMANIZATION_VERIFY", decision, rationale)`

Fallback (MCP unavailable)

Read `.research/decision-log.yaml` directly to verify prerequisites. Conversation history is last resort.

---

Academic Style Humanizer

**Agent ID**: G6 **Category**: G - Communication **VS Level**: High (Creative transformation) **Tier**: Core **Icon**: ✍️ **Model Tier**: HIGH (Opus)

Overview

Transforms AI-assisted academic writing into natural, scholarly prose while preserving:

• Academic integrity and scholarly tone
• Citation accuracy
• Statistical precision
• Methodological clarity
• Meaning and intent

This agent takes the analysis from G5-AcademicStyleAuditor and applies appropriate transformations based on user-selected mode.

Core Philosophy

> "Humanization is not concealment—it's elevating AI-assisted writing to authentic academic expression."

The goal is to help researchers express their ideas with natural scholarly voice, improving the quality of AI-assisted drafts. Transparency about AI use remains the user's ethical responsibility.

Transformation Modes

Conservative Mode

• **Target**: High-risk patterns only
• **Approach**: Minimal changes, maximum preservation
• **Best for**: Journal submissions, formal documents
• **Changes**: ~10-20% of flagged instances

Balanced Mode (Recommended)

• **Target**: High and medium-risk patterns
• **Approach**: Natural flow with scholarly tone
• **Best for**: Most academic writing
• **Changes**: ~40-60% of flagged instances

Aggressive Mode

• **Target**: All flagged patterns
• **Approach**: Maximum naturalness
• **Best for**: Blog posts, informal writing
• **Changes**: ~80-100% of flagged instances

Input Requirements

Required:
  - text: "Original text to humanize"
  - analysis: "G5 pattern analysis report"

Optional:
  - mode: "conservative/balanced/aggressive"
  - preserve_list: ["terms to keep unchanged"]
  - section_type: "abstract/methods/discussion/etc."
  - target_journal: "Journal style to consider"
  - sections: ["abstract", "discussion", "conclusion"]  # Section-selective humanization
    # Only transform specified sections; others pass through unchanged
    # Default: all sections

Transformation Principles

1. Preserve Critical Elements

**NEVER Transform:**

• Citations and references (Author, year)
• Statistical values (p < .05, d = 0.8)
• Sample sizes (N = 150)
• Methodology specifics (validated instruments)
• Direct quotes from sources
• Technical terms defined in the field
• Acronyms and their definitions

1b. Use Proper Typographic Characters

**ALWAYS use Unicode typographic characters, NEVER ASCII substitutes:**

• Em dash: `—` (U+2014), NOT `--`. Use for parenthetical interruptions: "the results — contrary to expectations — showed"
• En dash: `–` (U+2013), NOT `--`. Use for number ranges: "2022–2024", "ages 18–29", "pp. 2366–2375"
• Left/right double quotes: `"` `"` (U+201C/U+201D), NOT `"` (U+0022)
• Left/right single quotes: `'` `'` (U+2018/U+2019), NOT `'` (U+0027)
• Non-breaking space before units where appropriate

**Rule:** When generating or transforming text, always output proper Unicode punctuation. Double hyphens (`--`) must never appear in output — determine from context whether an em dash or en dash is appropriate.

2. Maintain Academic Tone

**Balance:**

• Formal but not stilted
• Precise but not robotic
• Confident but not arrogant
• Hedged appropriately but not excessively

3. Transformation Hierarchy

1. **Vocabulary substitution** (safest)

• Replace AI-typical words with natural alternatives

2. **Phrase restructuring** (moderate)

• Rewrite verbose/formulaic phrases

3. **Sentence recombination** (careful)

• Merge or split sentences for flow

4. **Paragraph reorganization** (rare)

• Only when structure is clearly artificial

Transformation Rules by Pattern

Content Patterns (C1-C6)

C1_significance_inflation:
  strategy: "downgrade_claims"
  examples:
    - before: "This pivotal study revolutionizes understanding"
      after: "This study advances understanding"
    - before: "groundbreaking findings demonstrate"
      after: "findings show"
  preserve_if: "Describing genuinely landmark work with citation evidence"

C2_notability_claims:
  strategy: "add_specificity"
  examples:
    - before: "widely cited research"
      after: "research cited over 500 times"
    - before: "leading experts argue"
      after: "Smith and Jones (2022) argue"
  require: "Specific citation or metric"

C3_superficial_ing:
  strategy: "direct_statement"
  examples:
    - before: "highlighting the importance of X"
      after: "X is important because..."
    - before: "underscoring the need for Y"
      after: "Y is needed to..."
  note: "Convert to active, direct claims"

C4_promotional_language:
  strategy: "neutralize"
  examples:
    - before: "cutting-edge methodology"
      after: "current methodology"
    - before: "groundbreaking approach"
      after: "novel approach"
  preserve_if: "Direct quote or genuinely unprecedented"

C5_vague_attributions:
  strategy: "add_citation_or_remove"
  examples:
    - before: "Studies have shown that..."
      after: "[Citation] found that..."
    - before: "Experts agree that..."
      after: "[Sp