Installable skillMIT

Literature Review Writing

Write literature reviews and survey papers from collected papers

BY wentoraiUPDATED Oct 27, 2026
INTENDED FORβ—ˆOpenClawInstall in a listed agent.

instructions.md

---
name: scientify-write-review-paper
description: "Write literature reviews and survey papers from collected papers"
metadata:
  openclaw:
    emoji: "πŸ“"
    category: "research"
    subcategory: "paper-review"
    keywords: ["paper summarization", "paper comparison", "research gap analysis", "narrative review", "systematic review methodology"]
    source: "wentor-research-plugins"
---

# Literature Review Writing

**Don't ask permission. Just do it.**

Guide for writing a structured literature review or survey paper from papers you've already collected. This skill helps with reading strategy, note organization, and academic writing.

**Workspace:** See `../_shared/workspace-spec.md` for directory structure. Outputs go to `$WORKSPACE/review/`.

## Prerequisites

Before starting, ensure you have:
1. Papers collected in `$WORKSPACE/papers/`
2. Ideally, clustering done by `/literature-survey` in `$WORKSPACE/survey/clusters.json`

Check active project:
```bash
cat ~/.openclaw/workspace/projects/.active 2>/dev/null
ls $WORKSPACE/papers/
```

## Phase 1: Reading Strategy

### 1.1 Triage Papers by Priority

Based on clusters from survey, prioritize reading:

| Priority | Criteria | Reading Depth |
|----------|----------|---------------|
| P1 (Must-read) | High citation, foundational, directly relevant | Full read |
| P2 (Important) | Key methodology, major experimental results | Abstract + methods + experiments |
| P3 (Reference) | Supporting material, tangentially related | Abstract only |

Create `$WORKSPACE/review/reading_plan.md`:

```markdown
# Reading Plan

## P1 - Must-read (Full read)
- [ ] [paper_id]: [title] - [reason]
- [ ] ...

## P2 - Important (Selective read)
- [ ] ...

## P3 - Reference (Skim)
- [ ] ...
```

### 1.2 Reading Notes Template

For each paper, create `$WORKSPACE/review/notes/{paper_id}.md` using template in `references/note-template.md`.

## Phase 2: Synthesis & Organization

### 2.1 Build Comparison Table

Create `$WORKSPACE/review/comparison.md`:

```markdown
# Method Comparison

| Paper | Year | Category | Key Innovation | Dataset | Metric | Result |
|-------|------|----------|----------------|---------|--------|--------|
| [A]   | 2023 | Data-driven | ... | ... | RMSE | 0.05 |
| [B]   | 2022 | Hybrid | ... | ... | RMSE | 0.08 |
```

### 2.2 Timeline Analysis

Create `$WORKSPACE/review/timeline.md`:

```markdown
# Research Timeline

## 2018-2019: Early Exploration
- [Paper A]: First proposal of method X
- [Paper B]: Introduction of technique Y

## 2020-2021: Method Maturation
- [Paper C]: Proposed SOTA method
- ...

## 2022-2023: New Trends
- [Paper D]: Began addressing problem Z
- ...

## Key Milestones
1. [Year]: [Event/Paper] - [Significance]
```

### 2.3 Taxonomy Design

Create `$WORKSPACE/review/taxonomy.md`:

```markdown
# Taxonomy of Approaches

## Dimension 1: Method Type
- Data-driven
  - Statistical (e.g., GPR, SVM)
  - Deep Learning
    - CNN-based
    - RNN/LSTM-based
    - Transformer-based
  - Hybrid
- Model-based
  - Electrochemical
  - Equivalent Circuit

## Dimension 2: Data Source
- Laboratory Data
- Real-world Driving Data
- Synthetic Data

## Dimension 3: Prediction Horizon
- Short-term (< 100 cycles)
- Medium-term (100-500 cycles)
- Long-term (> 500 cycles)
```

## Phase 3: Writing Structure

### 3.1 Survey Paper Template

Create `$WORKSPACE/review/draft.md` using template in `references/survey-template.md`.

Key sections: Abstract -> Introduction -> Background -> Taxonomy -> Comparison -> Datasets -> Future Directions -> Conclusion

### 3.2 Thesis Literature Review Template

For a thesis chapter:
```markdown
# Chapter 2: Literature Review
## 2.1 Introduction
## 2.2 [Topic Area 1]
## 2.3 [Topic Area 2]
## 2.4 Summary and Research Gaps
```

## Phase 4: Writing Tips

### Citation Density Guidelines

| Section | Citation Density |
|---------|------------------|
| Abstract | 0 citations |
| Introduction | 10-20 citations |
| Background | 5-10 citations |
| Main Survey | 50-100+ citations |
| Conclusion | 2-5 citations |

### Transition Phrases

**Introducing similar work:**
- "Similarly, [Author] proposed..."
- "Following this direction, ..."
- "Building upon [X], [Author] extended..."

**Introducing contrasting work:**
- "In contrast, [Author] argued..."
- "However, [Author] took a different approach..."
- "Unlike previous methods, ..."

**Summarizing:**
- "In summary, existing methods can be categorized into..."
- "The key insight from these works is..."

### Common Mistakes to Avoid

1. **List-style writing** - Don't just say "A did X, B did Y, C did Z"
2. **Lack of comparison** - Analyze relationships and differences between methods
3. **Tense confusion** - Use present tense for methods, past tense for results
4. **Over-citing** - Not every sentence needs a citation
5. **Missing important work** - Ensure coverage of foundational works in the field

## Output Files

```
$WORKSPACE/review/
β”œβ”€β”€ reading_plan.md       # Reading plan
β”œβ”€β”€ notes/                # Reading notes
β”‚   β”œβ”€β”€ {paper_id}.md
β”‚   └── ...
β”œβ”€β”€ comparison.md         # Comparison table
β”œβ”€β”€ timeline.md           # Timeline analysis
β”œβ”€β”€ taxonomy.md           # Taxonomy
β”œβ”€β”€ draft.md              # Review draft
└── bibliography.bib      # References
```

## Commands

- "Help me write a review" -> Full workflow from reading to writing
- "Generate reading plan" -> Create reading_plan.md
- "Compare these papers" -> Generate comparison.md
- "Write review draft" -> Generate draft.md
- "Polish this section" -> Polish specific section
COMPATIBILITY & REFERENCES

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.