Academic-Writing: Skill for Claude Code

Literature Review Rewrite Orchestrator

Rewrite Chapter 2 from scratch with deep extraction from ~20 cited papers. Uses a 5-phase pipeline with 7 specialized agents.

Arguments

• all — Run the full 5-phase pipeline (default)
• phase1 through phase5 — Run a specific phase (assumes prior phases completed)
• section N — Draft only section 2.N (for iterating on one section)

Workspace

All intermediate artifacts go in thesis/chapters/.litreview_workspace/. Create it at the start:

mkdir -p /Users/alessandro/Projects/Tesi/thesis/chapters/.litreview_workspace/drafts

Target Structure (~400-500 lines)

Chapter 2: Literature Review
├── Intro paragraph (preview sections using \Cref)
├── 2.1 Automation and the Labor Market
│   ├── 2.1.1 The Task-Based Framework
│   ├── 2.1.2 Labor Market Polarization
│   ├── 2.1.3 Task Displacement and Rent Dissipation
│   └── 2.1.4 Bridge paragraph → politics
├── 2.2 Economic Shocks and Political Attitudes
│   ├── 2.2.1 The China Shock and Political Consequences
│   ├── 2.2.2 Automation and Voting Behavior
│   └── 2.2.3 Methodological Parallels (NEW)
├── 2.3 Mechanisms: From Economic Disruption to Political Change
│   ├── 2.3.1 The Economic Insecurity Thesis
│   ├── 2.3.2 The Cultural Backlash Thesis
│   ├── 2.3.3 Status Threat and the 2016 Election
│   ├── 2.3.4 Psychological Foundations (DEEPENED)
│   └── 2.3.5 Identity Threat as Integrating Framework (NEW)
└── 2.4 Contribution of This Thesis

---

PHASE 1: RESEARCH (parallel)

Launch TWO agents in parallel using the Task tool:

Agent 1: Extraction Assembler (sonnet, read-only)

subagent_type: general-purpose
model: sonnet

Prompt:

You are assembling a structured evidence dossier for a thesis literature review rewrite.

Read ALL files in thesis/references/literature_analysis/paper_extractions/*.json (12 files), the concept graph at thesis/references/literature_analysis/concept_graph.json, and the mechanism extraction at thesis/references/literature_analysis/mechanism_extraction.json.

Produce TWO JSON files in thesis/chapters/.litreview_workspace/:

dossier.json — Organize ALL extracted evidence by the target section structure:

{
  "2.1.1_task_framework": {
    "papers": ["autor_levy_murnane_2003"],
    "key_quotes": [{"text": "...", "page": N, "bibkey": "autor2003skill"}],
    "claims": [{"claim": "...", "evidence": "...", "strength": "strong"}],
    "concept_edges": ["task_model → routine_manual", "task_model → substitution"],
    "constructs": [{"name": "...", "definition": "..."}]
  },
  "2.1.2_polarization": { ... },
  ...
}

gaps.json — Identify areas where extractions are thin and RAG queries would help:

{
  "gaps": [
    {
      "section": "2.1.3",
      "topic": "rent dissipation mechanism details",
      "query": "rent dissipation automation wage premium loss",
      "reason": "Extraction has claim but no detailed mechanism quote"
    },
    ...
  ]
}

Map evidence to these sections:

• 2.1.1: autor_levy_murnane_2003 (task framework, DOT, routine/nonroutine)
• 2.1.2: autor_dorn_2013 (polarization, U-shape, commuting zones, shift-share)
• 2.1.3: acemoglu_restrepo_2025 (task displacement, labor share, rent dissipation)
• 2.2.1: autor_dorn_hanson_majlesi_2020, colantone_stanig_2018 (China shock politics)
• 2.2.2: frey_berger_chen_2018, anelli_colantone_stanig_2019 (robots and voting)
• 2.2.3: methodological parallels (shift-share across studies, pseudo-panel)
• 2.3.1: economic insecurity evidence (Colantone & Stanig, Norris/Inglehart)
• 2.3.2: inglehart_norris_2016, norris_inglehart_2018 (cultural backlash)
• 2.3.3: mutz_2018, morgan_2018 (status threat debate)
• 2.3.4: osborne_2023, stenner_2005 (psychological mechanisms, dual-process)
• 2.3.5: manunta_2025, gidron_hall_2017 (identity threat integration)
• 2.4: contribution positioning

Include ALL verbatim quotes with page numbers. Be exhaustive — this is the evidence base for the entire chapter.