@thuanphan2208/paper-pilot 1.0.0

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

@@ -0,0 +1,125 @@
+---
+name: paper-write-method
+description: Write a draft Methodology section for a research paper via guided Q&A. Works across all academic majors and paper types. Use after completing paper:teach-method.
+license: MIT
+metadata:
+  author: claude-paper-skills
+  version: "2.0"
+---
+
+You are helping a student write their **Methodology** section. Read context, conduct Q&A, then write a draft tailored to the paper type.
+
+**English only** — Respond in English regardless of what language the user writes in.
+
+---
+
+## Step 1: Read Context
+
+Check `paper/context.yaml`. If missing, ask the user to run `/paper:plan` first. Use `paper_type` and `contribution_type` to determine which Q&A path to follow.
+
+**Format**: Read `writing_format` from context.yaml and apply throughout the draft:
+- `latex` → use `\section{}`, `\subsection{}`, `\cite{ref}`, math in `$...$` or `\begin{equation}`, pseudocode in `algorithm` environment
+- `word` → plain text with labelled headings, no LaTeX commands
+- `markdown` → use `##`, `**bold**`, code blocks for pseudocode
+
+- **empirical / new-method** → System design, components, algorithm
+- **review** → Search strategy, inclusion/exclusion criteria, screening process
+- **theoretical** → Framework components, theoretical grounding, application example
+
+---
+
+## Step 2: Q&A (one at a time, wait for each answer)
+
+### If empirical / new-method:
+
+Tell the user: "I'll ask about your system or method so I can write the Methodology. More detail = more reproducible draft."
+
+**Q1**: "Describe your system or method step by step. What is the input? What is the output? What happens in between?"
+*(Example: "Input: a URL string. Step 1: extract 48 features. Step 2: pass to Random Forest classifier. Output: phishing/legitimate label + confidence score.")*
+
+**Q2**: "How many main components or modules does your system have? What does each one do?"
+*(Example: "3 components — Feature Extractor (parses URL and HTML), Classifier (Random Forest), Decision Engine (applies 0.5 threshold to confidence score).")*
+
+**Q3**: "What algorithm, model, or technique do you use? With what configuration or parameters?"
+*(Example: "Random Forest: 100 trees, max depth 10, min samples per leaf 5. Hyperparameters chosen via 5-fold cross-validation.")*
+
+**Q4**: "Why did you choose this method/algorithm over alternatives?"
+*(Example: "Random Forest over SVM because it handles non-linear interactions and provides feature importance scores, which we use for our weighting mechanism.")*
+
+**Q5**: "Are there any formulas, pseudocode, or special technical details that need to be described?"
+*(Example: "Yes — the feature weighting formula: w_i = importance_i / Σimportance. Also pseudocode for the prediction pipeline.")*
+
+---
+
+### If review:
+
+**Q1**: "Which databases or sources did you search? (e.g., Google Scholar, Scopus, PubMed, IEEE Xplore)"
+
+**Q2**: "What keywords or search terms did you use?"
+
+**Q3**: "What were your inclusion and exclusion criteria? What made a paper qualify for review?"
+
+**Q4**: "How many papers did your search return initially? How many did you end up including after screening?"
+
+---
+
+### If theoretical / framework:
+
+**Q1**: "What existing theory, model, or principles does your framework build on?"
+
+**Q2**: "What are the main components or dimensions of your framework? How do they relate to each other?"
+
+**Q3**: "How would someone apply your framework? Walk me through an example."
+
+---
+
+## Step 3: Write Draft
+
+### If empirical / new-method:
+
+**Opening overview paragraph:**
+"Figure X illustrates the overall architecture of [system name]. Given [input] as input, the system performs [N] sequential steps: (1) [step 1], (2) [step 2], ..., producing [output] as output."
+
+**Subsection per component:**
+```
+3.1 [Component Name]
+Describe what it does, how it works, key parameters.
+Justify choices: "We employ X because Y."
+```
+
+**Algorithm / model:**
+- Existing algorithm: cite it and describe your configuration.
+- Novel algorithm: write numbered pseudocode steps.
+
+---
+
+### If review:
+
+**Search Strategy:**
+"We searched [databases] using the keywords [terms] for papers published between [years]. We included [criteria]. We excluded [criteria]."
+
+**Screening:**
+"An initial search returned [N] papers. After title and abstract screening, [N] papers remained. After full-text review, [N] papers were included for analysis."
+
+---
+
+### If theoretical:
+
+**Theoretical grounding:**
+"Our framework builds on [theory] [cite], which proposes [key idea]. We extend this to [domain] by [what you add or modify]."
+
+**Framework components:**
+Define each component clearly with a diagram description if applicable.
+
+**Application example:**
+Walk through how the framework applies to a concrete case.
+
+---
+
+## Step 4: Save and Recommend
+
+Ask: "Shall I save this draft to `paper/sections/03-methodology.md`?"
+
+Save and update context.yaml: `methodology: draft`.
+
+"Draft saved! Continue with `/paper:teach experiment` or review this section with `/paper:review method`."

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

@@ -0,0 +1,74 @@
+---
+name: paper-write-related
+description: Write a draft Related Work section for a research paper via guided Q&A. Works across all academic majors. Use after completing paper:teach-related.
+license: MIT
+metadata:
+  author: claude-paper-skills
+  version: "2.0"
+---
+
+You are helping a student write their **Related Work** section. Read context, conduct Q&A about papers they've read, then write a thematically organized draft.
+
+**English only** — Respond in English regardless of what language the user writes in.
+
+---
+
+## Step 1: Read Context
+
+Check `paper/context.yaml`. If missing, ask the user to run `/paper:plan` first. Use the topic and field to understand what kind of related work to expect.
+
+**Format**: Read `writing_format` from context.yaml and apply throughout the draft:
+- `latex` → use `\section{}`, `\subsection{}`, `\cite{ref}`, `\textbf{}`
+- `word` → plain text with labelled headings, no LaTeX commands
+- `markdown` → use `##`, `**bold**`, `-` for bullets
+
+---
+
+## Step 2: Q&A (one at a time, wait for each answer)
+
+Tell the user: "I need to know about the papers you've read to write your Related Work. Answer as best you can — if you haven't read many papers yet, I'll help you think through what to look for."
+
+**Q1**: "List 5–10 related papers you've read or know of. Format: [Author(s), year — brief description or method name]"
+*(Example: "Zhang et al. 2020 — SVM with URL features; Google Safe Browsing 2022 — blacklist API; Li et al. 2021 — LSTM on HTML content")*
+*(If few papers: "Search Google Scholar for '[your topic] survey' or '[your topic] review' filtered to last 5 years. Aim for at least 8 papers.")*
+
+**Q2**: "How would you group those papers into 2–3 themes or categories? What patterns do you see?"
+*(Example: "Group 1: blacklist-based — PhishTank, Safe Browsing. Group 2: ML with URL features — Zhang, Li, Kim. Group 3: deep learning on page content — Chen, Wang.")*
+*(If stuck: "Look at the method each paper uses. Papers using similar methods belong in the same group.")*
+
+**Q3**: "What are the main limitations of existing approaches in your area? Why don't they fully solve your problem?"
+*(Example: "Blacklist methods are too slow — update takes hours. ML methods use static feature sets that don't adapt to new attack patterns. Deep learning is too slow for real-time use.")*
+
+**Q4**: "What makes your approach different from the prior work you've reviewed?"
+*(Example: "Unlike prior ML approaches that use fixed features, we dynamically weight features by predictive power. Also the first study specifically on Vietnamese phishing campaigns.")*
+
+---
+
+## Step 3: Write Draft
+
+Structure the Related Work with 3 moves:
+
+**Opening sentence:**
+"Prior work on [topic] can be broadly categorized into [N] approaches: [theme 1], [theme 2], [and theme 3]."
+
+**For each theme (1–2 paragraphs):**
+- Name the approach and group the relevant papers together
+- Describe what they do at a general level
+- Critique their limitations using transition language:
+  - "While [these methods] achieve [result], they [limitation] because [reason]."
+  - "A common limitation of [theme] is [issue], which [consequence]."
+
+**Positioning paragraph (end of section):**
+"Unlike [prior approaches], our work [differentiator]. To the best of our knowledge, this is the first work to [novelty claim if applicable]."
+
+Adapt tone and framing to the user's field. In social sciences or humanities, positioning may be more about theoretical lens than technical novelty.
+
+---
+
+## Step 4: Save and Recommend
+
+Ask: "Shall I save this draft to `paper/sections/02-related-work.md`?"
+
+If yes, save and update context.yaml: `related_work: draft`.
+
+"Draft saved! Continue with `/paper:teach method` or review this section with `/paper:review related`."

package/.claude/skills/paper-write-results/SKILL.md

@@ -0,0 +1,131 @@
+---
+name: paper-write-results
+description: Write a draft Results section for a research paper via guided Q&A. Works across all academic majors and paper types. Use after completing paper:teach-results.
+license: MIT
+metadata:
+  author: claude-paper-skills
+  version: "2.0"
+---
+
+You are helping a student write their **Results** section. Push the user to go beyond reporting numbers — analysis and interpretation are what make results compelling.
+
+**English only** — Respond in English regardless of what language the user writes in.
+
+---
+
+## Step 1: Read Context
+
+Check `paper/context.yaml`. Use `paper_type` to determine the right approach.
+
+**Format**: Read `writing_format` from context.yaml and apply throughout the draft:
+- `latex` → use `\section{}`, `\subsection{}`, tables in `tabular`, bold with `\textbf{}`, `\cite{ref}`
+- `word` → plain text with labelled headings, ASCII tables or described table structure
+- `markdown` → use `##`, tables in Markdown pipe syntax, `**bold**`
+
+- **empirical / new-method** → Comparison table + analysis + ablation
+- **review** → Synthesis of findings + taxonomy + open challenges
+- **theoretical** → Framework validation, case application, discussion of implications
+
+Also check if `paper/sections/04-experiments.md` exists — read it if available to understand baselines and metrics already defined.
+
+---
+
+## Step 2: Q&A (one at a time, wait for each answer)
+
+### If empirical / new-method:
+
+Tell the user: "I need your experimental results to write this section."
+
+**Q1**: "What are your main results? List performance numbers for your method and all baselines across each metric."
+*(Example: "PhishGuard: Acc=97.4%, P=97.1%, R=97.7%, F1=97.4%. SVM-URL: Acc=92.1%, F1=92.1%. PhishTank blacklist: Acc=78.3%, F1=76.2%.")*
+*(If results are worse than baselines: "Share the numbers honestly — we'll handle how to write about it.")*
+
+**Q2**: "Does your method outperform the baselines? By how much on the primary metric?"
+*(Example: "Yes — beats SVM-URL by 5.3% F1, and blacklist by 21.2%.")*
+*(If no: "By how much does it fall short? On which metric? Does it win on any specific metric or subset?")*
+
+**Q3**: "Why do you think your method performs better? What design decision drives that improvement?"
+*(Example: "The feature weighting mechanism — it assigns more importance to discriminative features like URL length, reducing noise from weaker signals.")*
+*(If unsure: "Look at your ablation results — which component removal caused the biggest performance drop?")*
+
+**Q4**: "Were there any surprising findings — a baseline that performed unexpectedly, or a metric that diverged from expectations?"
+*(Example: "PhishGuard-NoWeight beat SVM-URL even without feature weighting — suggesting Random Forest's ensemble nature itself provides robustness.")*
+
+**Q5**: "Do you have ablation study results? (performance when each component is removed)"
+*(Example: "Yes — removing URL features drops F1 by 8.2%; domain features by 3.3%; HTML features by 1.2%.")*
+*(If no ablation: "That's fine — we'll note it as a limitation of the study.")*
+
+---
+
+### If review:
+
+**Q1**: "What are the main patterns or trends you found across the reviewed literature?"
+
+**Q2**: "Were there consistent findings across papers, or significant disagreements?"
+
+**Q3**: "What gaps or open challenges did the literature reveal?"
+
+---
+
+### If theoretical:
+
+**Q1**: "How did you validate or demonstrate your framework? What case or example did you use?"
+
+**Q2**: "What did the framework reveal or explain that prior work could not?"
+
+---
+
+## Step 3: Write Draft
+
+### If empirical / new-method:
+
+**[Section N].1 Main Results**
+
+Format a comparison table:
+```
+Table 1: [Metric] comparison on [dataset]
+
+Method              [Metric 1]  [Metric 2]  [Metric 3]
+──────────────────────────────────────────────────────────
+[Baseline 1]         X%          X%          X%
+[Baseline 2]         X%          X%          X%
+[Your method]      **X%**      **X%**      **X%**
+──────────────────────────────────────────────────────────
+Bold = best result
+```
+
+Analysis paragraph (2–3 sentences — NOT just restating numbers):
+"[Your method] achieves [best metric], outperforming [best baseline] by [X] points. This improvement can be attributed to [reason from Q3]. Notably, [any surprising finding from Q4]."
+
+**[Section N].2 Ablation Study** (if the user has data):
+Table with each component removed + performance delta.
+"The ablation results confirm that [most impactful component] contributes most significantly, with a [X%] drop when removed."
+
+**[Section N].3 Discussion** (optional):
+Broader interpretation — what do these results mean for the field? Any limitations of the results themselves?
+
+---
+
+### If review:
+
+**Synthesis:**
+Write thematic paragraphs summarizing what the literature found across each category. Highlight agreements, disagreements, and evolving trends.
+
+**Open Challenges:**
+"Despite [progress], several challenges remain unaddressed: [challenge 1], [challenge 2], [challenge 3]."
+
+---
+
+### If theoretical:
+
+Describe how the framework was applied, what it revealed, and how it compares to prior conceptualizations.
+
+---
+
+## Step 4: Save and Recommend
+
+Ask: "Shall I save this draft to `paper/sections/05-results.md`?"
+
+Save and update context.yaml: `results: draft`.
+
+"Draft saved! Continue with `/paper:teach conclusion` or review this section with `/paper:review results`."

package/README.md

@@ -0,0 +1,2 @@
+# Claude_Skills-for-Research
+A collection of Claude Code skills for writing CS/IT research papers — from brainstorming to draft. Includes teach + write skills for every section (Introduction, Related Work, Methodology, Experiments, Results, Conclusion, Abstract), with a complete phishing detection demo paper.

package/bin/cli.js

@@ -0,0 +1,46 @@
+#!/usr/bin/env node
+'use strict';
+
+const { init } = require('../src/init');
+
+const args = process.argv.slice(2);
+const command = args[0];
+const flags = args.slice(1);
+const dryRun = flags.includes('--dry-run');
+
+const BOLD = '\x1b[1m';
+const RESET = '\x1b[0m';
+
+switch (command) {
+  case 'init':
+    init({ dryRun });
+    break;
+
+  case '--version':
+  case '-v': {
+    const pkg = require('../package.json');
+    console.log(pkg.version);
+    break;
+  }
+
+  case '--help':
+  case '-h':
+  case undefined:
+    console.log(`
+${BOLD}paper-pilot${RESET} — Install Claude Code paper-writing skills into your project
+
+Usage:
+  npx paper-pilot init           Install skills into current directory
+  npx paper-pilot init --dry-run Preview what would be installed
+  npx paper-pilot --version      Show version
+  npx paper-pilot --help         Show this help
+
+After init, open Claude Code and run /paper:explore to get started.
+`);
+    break;
+
+  default:
+    console.error(`Unknown command: ${command}`);
+    console.error('Run "npx paper-pilot --help" for usage.');
+    process.exit(1);
+}

package/package.json

@@ -0,0 +1,29 @@
+{
+  "name": "@thuanphan2208/paper-pilot",
+  "version": "1.0.0",
+  "description": "Install Claude Code paper-writing skills into any project with one command.",
+  "keywords": ["claude", "claude-code", "research", "paper", "skills", "academic"],
+  "homepage": "https://github.com/ThuanPne/Claude_Skills-for-Research",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/ThuanPne/Claude_Skills-for-Research.git"
+  },
+  "license": "MIT",
+  "bin": {
+    "paper-pilot": "./bin/cli.js"
+  },
+  "files": [
+    "bin/",
+    "src/",
+    ".claude/"
+  ],
+  "scripts": {
+    "test": "node bin/cli.js init --dry-run"
+  },
+  "engines": {
+    "node": ">=18.0.0"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}

package/src/init.js

@@ -0,0 +1,111 @@
+const fs = require('fs');
+const path = require('path');
+
+const GREEN = '\x1b[32m';
+const YELLOW = '\x1b[33m';
+const CYAN = '\x1b[36m';
+const BOLD = '\x1b[1m';
+const RESET = '\x1b[0m';
+
+function log(msg) { process.stdout.write(msg + '\n'); }
+function success(msg) { log(`${GREEN}✓${RESET} ${msg}`); }
+function skip(msg) { log(`${YELLOW}─${RESET} ${msg} (already exists, skipped)`); }
+function info(msg) { log(`${CYAN}  ${msg}${RESET}`); }
+
+function copyDirMerge(src, dest, dryRun = false) {
+  const entries = fs.readdirSync(src, { withFileTypes: true });
+  let copied = 0;
+  let skipped = 0;
+
+  for (const entry of entries) {
+    const srcPath = path.join(src, entry.name);
+    const destPath = path.join(dest, entry.name);
+
+    if (entry.isDirectory()) {
+      if (!dryRun && !fs.existsSync(destPath)) {
+        fs.mkdirSync(destPath, { recursive: true });
+      }
+      const result = copyDirMerge(srcPath, destPath, dryRun);
+      copied += result.copied;
+      skipped += result.skipped;
+    } else {
+      if (fs.existsSync(destPath)) {
+        skip(path.relative(process.cwd(), destPath));
+        skipped++;
+      } else {
+        if (!dryRun) {
+          fs.mkdirSync(dest, { recursive: true });
+          fs.copyFileSync(srcPath, destPath);
+        }
+        success(path.relative(process.cwd(), destPath));
+        copied++;
+      }
+    }
+  }
+
+  return { copied, skipped };
+}
+
+function ensureDir(dirPath, dryRun = false) {
+  if (fs.existsSync(dirPath)) {
+    skip(path.relative(process.cwd(), dirPath) + '/');
+  } else {
+    if (!dryRun) fs.mkdirSync(dirPath, { recursive: true });
+    success(path.relative(process.cwd(), dirPath) + '/');
+  }
+}
+
+function init({ dryRun = false } = {}) {
+  const packageRoot = path.join(__dirname, '..');
+  const destRoot = process.cwd();
+
+  log('');
+  log(`${BOLD}paper-pilot init${RESET}${dryRun ? ' (dry run)' : ''}`);
+  log('─'.repeat(50));
+
+  // 1. Copy .claude/ (skills + commands)
+  const claudeSrc = path.join(packageRoot, '.claude');
+  const claudeDest = path.join(destRoot, '.claude');
+
+  if (!fs.existsSync(claudeSrc)) {
+    log(`\x1b[31mError: .claude/ not found in package. Reinstall paper-pilot.\x1b[0m`);
+    process.exit(1);
+  }
+
+  log(`\nInstalling skills into ${CYAN}.claude/${RESET}`);
+  const { copied, skipped } = copyDirMerge(claudeSrc, claudeDest, dryRun);
+
+  // 2. Create paper/sections/ directory
+  log(`\nCreating paper workspace`);
+  ensureDir(path.join(destRoot, 'paper'), dryRun);
+  ensureDir(path.join(destRoot, 'paper', 'sections'), dryRun);
+
+  // 3. Create paper/notes/.gitkeep if not present
+  const notesDir = path.join(destRoot, 'paper', 'notes');
+  const notesKeep = path.join(notesDir, '.gitkeep');
+  if (!fs.existsSync(notesKeep)) {
+    if (!dryRun) {
+      fs.mkdirSync(notesDir, { recursive: true });
+      fs.writeFileSync(notesKeep, '');
+    }
+    success('paper/notes/');
+  } else {
+    skip('paper/notes/');
+  }
+
+  // 4. Summary
+  log('');
+  log('─'.repeat(50));
+  log(`${copied} file(s) installed, ${skipped} skipped`);
+  log('');
+  log(`${BOLD}You're ready! Open Claude Code and run:${RESET}`);
+  log('');
+  info('/paper:explore   ← find your research topic');
+  info('/paper:plan      ← plan your paper structure');
+  info('/paper:teach     ← learn how to write each section');
+  info('/paper:write     ← draft each section with AI guidance');
+  info('/paper:review    ← get feedback on your drafts');
+  log('');
+}
+
+module.exports = { init };