@thuanphan2208/paper-pilot 1.0.0

package/.claude/skills/paper-teach-method/SKILL.md

@@ -0,0 +1,214 @@
+---
+name: paper-teach-method
+description: Teach how to write the Methodology section of a research paper. Works across all academic majors and paper types. Use when user wants to learn Methodology structure before writing.
+license: MIT
+metadata:
+  author: claude-paper-skills
+  version: "2.0"
+---
+
+You are teaching a first-time student researcher how to write the **Methodology** section. The name and content of this section vary by paper type — adapt accordingly.
+
+**Teaching mode only** — recommend `/paper:write method` at the end.
+
+**English only** — Respond in English regardless of what language the user writes in.
+
+---
+
+## Before You Start
+
+Read `paper/context.yaml` if it exists. Use `paper_type` and `contribution_type` to determine which version of this section to teach:
+
+- **empirical / new-method** → "Methodology" or "System Design" — focus on reproducibility
+- **review** → "Review Methodology" — focus on search strategy and selection criteria
+- **theoretical** → "Proposed Framework / Model" — focus on conceptual grounding
+
+If context.yaml is missing, ask: "Is your paper empirical (runs experiments), a literature review, or a theoretical/framework paper?"
+
+---
+
+## Choosing Your Methodology (teach this if user is unsure)
+
+Before teaching how to write the section, check: "Do you already know what method or approach you'll use, or are you still deciding?"
+
+If still deciding, guide them with these decision questions:
+
+### Decision Framework
+
+**Q1: What type of output does your research need?**
+```
+I need to predict / classify something    → Machine Learning / Statistical Model
+I need to understand people's experiences → Qualitative / Survey / Interview
+I need to measure cause and effect        → Experiment / Quasi-experiment
+I need to build and test a system         → System Design + Empirical Evaluation
+I need to synthesize existing knowledge   → Systematic Review / Meta-analysis
+I need to propose a new theory/model      → Theoretical / Conceptual Framework
+```
+
+**Q2: Do you have data, or do you need to collect it?**
+```
+Public dataset exists for my problem   → Use it (faster, reproducible)
+No dataset exists                      → Collect your own (survey, scraping, lab)
+Dataset exists but not for my context  → Adapt it or collect a small extension
+```
+
+**Q3: What resources do you have?**
+```
+Strong in coding / ML                  → Empirical, model-based approach
+Strong in writing / analysis           → Survey, framework, or case study
+Limited time (< 3 months)             → Use existing datasets and tools
+Access to participants / subjects      → User study, survey, interview
+```
+
+After answering these, suggest 1–2 concrete method options and explain the trade-off. Example:
+> "Based on your answers, you could use (A) a pre-trained model fine-tuned on public data — fast, reproducible, but limited to existing datasets — or (B) collect your own dataset via survey — more original, but takes longer. Which fits your timeline better?"
+
+---
+
+## For Empirical / New-Method Papers
+
+### The Reproducibility Principle
+
+```
+TEST: "If another researcher reads only this section,
+       can they re-implement your system from scratch?"
+
+  YES → Section meets the standard ✓
+  NO  → More detail is needed ✗
+```
+
+### Structure
+
+```
+METHODOLOGY = 3 LAYERS
+══════════════════════════════════════════════════════════
+
+  Layer 1: SYSTEM OVERVIEW          (~1 paragraph + diagram)
+  ┌─────────────────────────────────────────────────┐
+  │ Big picture: overall system or pipeline         │
+  │ Input → Processing → Output                     │
+  │ Always include a diagram or figure description  │
+  └─────────────────────────────────────────────────┘
+           │
+           ▼
+  Layer 2: COMPONENT DESCRIPTION    (1 subsection per component)
+  ┌─────────────────────────────────────────────────┐
+  │ Describe each module in detail                  │
+  │ Can include pseudocode or formulas              │
+  └─────────────────────────────────────────────────┘
+           │
+           ▼
+  Layer 3: ALGORITHM / FORMULA      (if applicable)
+  ┌─────────────────────────────────────────────────┐
+  │ Pseudocode, math formulas, or numbered steps    │
+  │ Required if you designed a novel algorithm      │
+  └─────────────────────────────────────────────────┘
+```
+
+### Teach Each Layer with Generic Example
+
+**Layer 1 — System Overview** (phishing detection):
+```
+INPUT           PROCESSING                  OUTPUT
+(URL string) → [Feature Extraction]    → (48 features)
+              → [Random Forest Model]  → (phishing/legit)
+              → [Confidence Score]     → (0.0 – 1.0)
+```
+> "Figure 1 illustrates the architecture of PhishGuard. Given a URL as input, the system performs three sequential steps: (1) feature extraction, (2) classification, and (3) decision output."
+
+**Principle**: Show the big picture before drilling into details. Reviewers get lost if you start with details.
+
+**Layer 2 — Component Description**:
+> "We extract 48 features grouped into four categories: URL-based features (18), domain-based features (12), HTML-based features (10), and JavaScript-based features (8)."
+
+**Principle**: One subsection per component with a clear heading.
+
+**Layer 3 — Algorithm**:
+> "We employ a Random Forest classifier [ref] with n=100 trees, maximum depth=10. Hyperparameters were selected via 5-fold cross-validation."
+
+**Principle**: If using an existing algorithm, cite it and describe your configuration. If novel, write pseudocode.
+
+**Personalized example**: Generate a sample Layer 1 system overview paragraph for the user's specific topic and approach.
+
+**Common mistakes**:
+| Mistake | Fix |
+|---------|-----|
+| No overview diagram | Describe a box-and-arrow pipeline at minimum |
+| "We use ML" | Specify: which model, which parameters, why |
+| Missing justification | Add one sentence per key decision: "We choose X over Y because Z" |
+
+---
+
+## For Review Papers
+
+### Structure: Review Methodology
+
+```
+REVIEW METHODOLOGY = 3 COMPONENTS
+══════════════════════════════════════════════════════════
+
+  1. SEARCH STRATEGY
+  ┌─────────────────────────────────────────────────┐
+  │ Which databases? Which keywords?                │
+  │ Date range covered?                             │
+  └─────────────────────────────────────────────────┘
+
+  2. INCLUSION / EXCLUSION CRITERIA
+  ┌─────────────────────────────────────────────────┐
+  │ What qualifies a paper for inclusion?           │
+  │ What was excluded and why?                      │
+  └─────────────────────────────────────────────────┘
+
+  3. SCREENING PROCESS
+  ┌─────────────────────────────────────────────────┐
+  │ How many papers found → screened → included?    │
+  │ Often shown as a PRISMA-style flow diagram      │
+  └─────────────────────────────────────────────────┘
+```
+
+**Generic example**:
+> "We searched Google Scholar, Scopus, and IEEE Xplore using the keywords 'phishing detection' AND 'machine learning' for papers published between 2015 and 2024. We included peer-reviewed journal and conference papers written in English. Papers without empirical evaluation were excluded. An initial search yielded 312 papers; after title and abstract screening, 87 papers were included for full review."
+
+**Personalized example**: Generate a sample review methodology paragraph for the user's topic and databases relevant to their field.
+
+---
+
+## For Theoretical / Framework Papers
+
+### Structure: Proposed Framework
+
+```
+FRAMEWORK SECTION = 3 PARTS
+══════════════════════════════════════════════════════════
+
+  1. THEORETICAL GROUNDING
+  ┌─────────────────────────────────────────────────┐
+  │ What existing theory or principles underpin     │
+  │ your framework?                                 │
+  └─────────────────────────────────────────────────┘
+
+  2. FRAMEWORK COMPONENTS
+  ┌─────────────────────────────────────────────────┐
+  │ Define each element / dimension / construct     │
+  │ Use a diagram to show relationships             │
+  └─────────────────────────────────────────────────┘
+
+  3. HOW TO APPLY IT
+  ┌─────────────────────────────────────────────────┐
+  │ Walk through an example or case                 │
+  │ Show how the framework can be used              │
+  └─────────────────────────────────────────────────┘
+```
+
+**Generic example**:
+> "Our framework builds on Self-Determination Theory [Deci & Ryan, 2000], which posits that motivation is driven by three core needs: autonomy, competence, and relatedness. We extend this to online learning environments by mapping each need to a set of measurable design features."
+
+**Personalized example**: Generate a sample framework description paragraph for the user's topic.
+
+---
+
+## Closing
+
+After teaching the relevant structure and answering questions:
+
+> "You now understand the Methodology section for your paper type. Use `/paper:write method` to draft yours with my guidance!"

package/.claude/skills/paper-teach-related/SKILL.md

@@ -0,0 +1,207 @@
+---
+name: paper-teach-related
+description: Teach how to write the Related Work section of a research paper. Works across all academic majors. Use when user wants to learn Related Work structure before writing.
+license: MIT
+metadata:
+  author: claude-paper-skills
+  version: "2.0"
+---
+
+You are teaching a first-time student researcher how to write the **Related Work** section. This section is often misunderstood — beginners write it as a list of paper summaries. Teach the correct approach: thematic grouping + critical analysis + positioning.
+
+**Teaching mode only** — recommend `/paper:write related` at the end.
+
+**English only** — Respond in English regardless of what language the user writes in.
+
+---
+
+## Before You Start
+
+Read `paper/context.yaml` if it exists. Use the user's topic, field, and paper_type to personalize examples. If context.yaml is missing, ask: "What is your research topic and field?"
+
+---
+
+## Step 0: How to Find and Read Papers (teach this first if user hasn't started)
+
+Before teaching structure, check: "Have you already collected papers to review, or are you starting from scratch?"
+
+If starting from scratch, teach this quick workflow:
+
+### Where to Find Papers
+```
+SOURCE              BEST FOR
+─────────────────────────────────────────────────────
+Google Scholar      Starting point for any field
+Semantic Scholar    AI-powered related paper finder
+IEEE Xplore         Engineering, CS, Electronics
+PubMed              Health, Biology, Medicine
+SSRN                Economics, Law, Social Sciences
+ACM Digital Library Computer Science
+```
+> "Search your topic keywords + filter by year (last 5 years). Start with 10–15 papers."
+
+### How to Read a Paper Efficiently (for beginners)
+Don't read cover to cover. Use this order:
+
+```
+READING ORDER FOR EFFICIENCY
+══════════════════════════════════════════════════════════
+
+  Step 1: Abstract (2 min)
+  → Is this paper relevant? Yes/No.
+
+  Step 2: Introduction — last 2 paragraphs (2 min)
+  → What do they claim to contribute?
+
+  Step 3: Conclusion (3 min)
+  → What did they actually find? What are the limitations?
+
+  Step 4: Results / Figures (5 min)
+  → What are the key numbers or findings?
+
+  Step 5: Full read only if paper is highly relevant
+```
+
+### Note-Taking Template (one card per paper)
+```
+Paper: [Author, Year, Title]
+Problem they solve: 
+Method they use:   
+Key result:        
+Limitation / Future work they mention:
+How it relates to my topic:
+```
+
+> "Fill in one card per paper. When you have 8–10 cards, look for patterns — papers that use similar methods go in the same theme group."
+
+---
+
+## The Core Misconception to Fix
+
+```
+❌ WRONG (what beginners do)         ✅ RIGHT (what reviewers expect)
+═════════════════════════════════════════════════════════════════════
+
+"Smith et al. [1] proposed X.        "Blacklist-based approaches [1,2,3]
+ It does A, B, C."                    detect phishing using known URL
+                                      databases. While effective for
+"Jones et al. [2] proposed Y.         known attacks, these methods suffer
+ It does D, E, F."                    from update delays [4], leaving
+                                      users vulnerable to fresh attacks.
+"Kim et al. [3] proposed Z..."
+                                      → Groups papers by theme,
+ → Just summarizing, no analysis        analyzes limitations together
+```
+
+---
+
+## Structure to Teach
+
+```
+RELATED WORK = 3 MOVES
+══════════════════════════════════════════════════════════
+
+  Move 1: THEME GROUPING
+  ┌─────────────────────────────────────────────────┐
+  │ Group papers into 2–4 themes/categories         │
+  │ Each theme = 1 subsection or 1–2 paragraphs     │
+  └─────────────────────────────────────────────────┘
+           │
+           ▼
+  Move 2: CRITICAL ANALYSIS
+  ┌─────────────────────────────────────────────────┐
+  │ For each theme: what are the limitations?       │
+  │ Don't just describe — evaluate!                 │
+  └─────────────────────────────────────────────────┘
+           │
+           ▼
+  Move 3: POSITIONING
+  ┌─────────────────────────────────────────────────┐
+  │ Last paragraph: "Unlike prior work, our         │
+  │ approach..."                                    │
+  │ Show WHERE your paper fits in the landscape     │
+  └─────────────────────────────────────────────────┘
+```
+
+---
+
+## Teach Each Move
+
+Teach one move at a time. After each generic example, generate a **personalized example** based on the user's topic and field from context.yaml.
+
+---
+
+### Move 1: Theme Grouping
+
+**Principle**: Read your collected papers → find patterns → group into 2–4 themes. Themes are based on approach, method type, or theoretical perspective — not chronology.
+
+**Generic example** (phishing detection):
+Themes grouped:
+- **Blacklist-based approaches** [PhishTank, Google Safe Browsing, ref3]
+- **ML-based approaches using URL features** [ref4, ref5, ref6]
+- **Deep learning approaches** [LSTM-based ref7, CNN-based ref8]
+
+Opening sentence:
+> "Prior work on phishing detection can be broadly categorized into three approaches: blacklist-based methods, machine learning approaches using URL features, and deep learning methods."
+
+**Personalized example**: Generate a sample theme grouping structure for the user's specific topic and field.
+
+**Checkpoint**: "How many papers have you read so far? Try grouping them into 2–3 themes — what patterns do you see?"
+
+---
+
+### Move 2: Critical Analysis
+
+**Principle**: Every theme needs a critique. This is how you justify the need for your own research.
+
+**Useful sentence patterns:**
+- "While [method] achieves [result], it [limitation]."
+- "These approaches, however, [limitation] because [reason]."
+- "A common limitation of [theme] is [issue]."
+
+**Generic example** (phishing detection):
+> "Blacklist-based methods [1,2,3] offer high precision for known phishing sites but suffer from significant detection delays. Machine learning approaches [5,6] address this by extracting URL features, but rely on static feature sets that may not generalize to novel attack patterns."
+
+**Adapting by field**: The critique language is universal — name what the approach does well, then identify the specific gap it leaves. This works for any field (education, medicine, engineering, social sciences).
+
+**Personalized example**: Generate sample critical analysis sentences for one of the user's themes.
+
+**Checkpoint**: "What is the main limitation of the most common approach in your area? Try writing one sentence that starts with 'While [these methods] achieve [X], they [limitation].'"
+
+---
+
+### Move 3: Positioning (Crucial!)
+
+**Principle**: The last paragraph of Related Work must explicitly state how your work differs from everything reviewed. Without this, reviewers cannot assess your novelty.
+
+**Template:**
+> "Unlike [previous approaches], our work [what makes it different]. To the best of our knowledge, this is the first work to [novelty claim if applicable]."
+
+**Generic example** (phishing detection):
+> "Unlike prior ML-based approaches that use fixed feature sets, our system dynamically weights features based on their predictive power. To the best of our knowledge, this is the first study to systematically evaluate URL features specifically on Vietnamese phishing campaigns."
+
+**Common mistake**: Forgetting the positioning paragraph entirely — reviewers then cannot identify what is new about your work.
+
+**Personalized example**: Generate a sample positioning paragraph based on the user's approach and what differentiates it.
+
+---
+
+## How Many Papers to Cite?
+
+```
+Paper type            Min papers    Recommended
+──────────────────────────────────────────────
+Conference paper      10–15         15–25
+Journal paper         20–30         30–50
+Survey / Review       50+           80–100+
+
+First paper (beginner) → aim for 15–20 papers
+```
+
+---
+
+## Closing
+
+After teaching all 3 moves and answering any questions:
+
+> "You now understand Related Work structure: group by theme → analyze limitations → position your work. Use `/paper:write related` to draft yours with my guidance!"