scientify 1.2.0 → 1.2.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "scientify",
-  "version": "1.2.0",
+  "version": "1.2.1",
   "description": "Scientify - AI-powered research workflow automation for OpenClaw. Includes idea generation, literature review, research pipeline skills, and arxiv tool.",
   "type": "module",
   "main": "dist/index.js",

package/skills/idea-generation/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: idea-generation
-description: "Generate innovative research ideas from a topic. SEARCHES arXiv/GitHub automatically, downloads papers, analyzes literature, and outputs 5 novel research ideas with arXiv citations. Use for: 找研究方向, 生成创新点, find research gaps, propose new methods. NOT for summarizing existing papers (use literature-review instead)."
+description: "Generate innovative research ideas from a topic. SEARCHES arXiv/GitHub automatically, downloads papers, analyzes literature, outputs 5 novel ideas with citations. Use for: 找研究方向, 生成创新点, find research gaps, propose new methods. NOT for: summarizing papers (use /write-review-paper), literature survey (use /literature-survey)."
 metadata:
   {
     "openclaw":
@@ -93,67 +93,14 @@ End-to-end workflow for generating innovative research ideas from a research top
 
 ## Step 0: Auto Project Management (REQUIRED)
 
-**Agent autonomously manages projects. DO NOT ask user for confirmation.**
+**Autonomous - DO NOT ask user for confirmation.**
 
-### 0.1 Extract Topic from User Query
+1. Extract topic from user query → convert to kebab-case ID
+2. Check `~/.openclaw/workspace/projects/` for existing match
+3. Use existing or create new: `mkdir -p $PROJECT_ID/{papers,repos,ideas}`
+4. Update `.active` file and set `$WORKSPACE` path
 
-Analyze the user's message to identify the research topic. Examples:
-- "帮我调研文本摘要方法" → topic: `text-summarization`
-- "推荐系统的深度学习方法" → topic: `rec-deep-learning`
-- "transformer attention optimization" → topic: `transformer-attention`
-
-Convert to kebab-case ID: lowercase, spaces/special chars → hyphens.
-
-### 0.2 Check Existing Projects
-
-```bash
-ls ~/.openclaw/workspace/projects/ 2>/dev/null | grep -v "^\.active$"
-```
-
-Read each `project.json` to check if topic matches:
-```bash
-cat ~/.openclaw/workspace/projects/*/project.json 2>/dev/null
-```
-
-### 0.3 Select or Create Project
-
-**If matching project exists**: Use it, update `.active`
-```bash
-echo "{project_id}" > ~/.openclaw/workspace/projects/.active
-```
-
-**If no match**: Create new project
-```bash
-PROJECT_ID="{topic-as-kebab-case}"
-mkdir -p ~/.openclaw/workspace/projects/$PROJECT_ID/{papers,repos,ideas}
-echo "$PROJECT_ID" > ~/.openclaw/workspace/projects/.active
-
-# Create project.json
-cat > ~/.openclaw/workspace/projects/$PROJECT_ID/project.json << 'EOF'
-{
-  "id": "{project_id}",
-  "name": "{Human readable name}",
-  "created": "{ISO date}",
-  "topics": ["{keyword1}", "{keyword2}"]
-}
-EOF
-```
-
-### 0.4 Set Working Paths
-
-After project selection, ALL subsequent paths use:
-```
-WORKSPACE=~/.openclaw/workspace/projects/{project_id}
-$WORKSPACE/task.json
-$WORKSPACE/search_results.md
-$WORKSPACE/papers/
-$WORKSPACE/repos/
-$WORKSPACE/ideas/
-$WORKSPACE/prepare_res.md
-```
-
-**Log project selection** (inform user briefly):
-> 📁 Using project: `{project_id}` ({new/existing})
+> 📁 Using project: `{project_id}` (new/existing)
 
 ---
 
@@ -415,55 +362,7 @@ For each paper in `papers/`:
 2. Extract: core contribution, math formulas, limitations, future work
 3. Note connections to other papers
 
-**⚠️ Handling Long Papers (>50KB or >15k tokens):**
-
-If a .tex file is too long to read in one pass:
-
-1. **First pass - Structure scan:**
-   ```bash
-   # List all .tex files and their sizes
-   ls -la $WORKSPACE/papers/{arxiv_id}/
-   # Check line count
-   wc -l $WORKSPACE/papers/{arxiv_id}/*.tex
-   ```
-
-2. **Chunked reading strategy:**
-   - Read `abstract` section first (usually in main.tex, first 200 lines)
-   - Read `\section{Introduction}` or `\section{Method}` separately
-   - Read `\section{Experiments}` or `\section{Results}` separately
-   - Read `\section{Conclusion}` and `\section{Related Work}` last
-
-   Use the Read tool with `offset` and `limit` parameters:
-   ```
-   Tool: Read
-   Arguments:
-     file_path: "$WORKSPACE/papers/2404.04429/main.tex"
-     offset: 1
-     limit: 500    # First 500 lines (abstract + intro)
-   ```
-
-   Then continue:
-   ```
-   Tool: Read
-   Arguments:
-     file_path: "$WORKSPACE/papers/2404.04429/main.tex"
-     offset: 500
-     limit: 500    # Lines 500-1000 (method section)
-   ```
-
-3. **Priority sections for idea generation:**
-   | Priority | Section | Why |
-   |----------|---------|-----|
-   | 1 | Abstract | Core contribution |
-   | 2 | Method/Approach | Technical details, formulas |
-   | 3 | Experiments | What works, what doesn't |
-   | 4 | Conclusion/Future Work | Limitations, open problems |
-   | 5 | Related Work | Connections to other papers |
-
-4. **Skip if context-limited:**
-   - Appendix (proofs, supplementary)
-   - Acknowledgments
-   - Detailed hyperparameter tables
+**Long Papers (>50KB):** See `references/reading-long-papers.md` for chunked reading strategy.
 
 For each repo in `repos/`:
 1. Understand structure: `gen_code_tree_structure` equivalent
@@ -478,38 +377,15 @@ Look for:
 - Scalability issues
 - Assumptions that could be relaxed
 
-### 5.3 Generate Idea 1
-
-Create `$WORKSPACE/ideas/idea_1.md` using the template in `references/idea-template.md`.
+### 5.3 Generate 5 Ideas
 
-**MUST include (with actual citations from your research):**
-- One-line summary
-- Challenges addressed
-- **Existing methods & limitations (cite specific papers by arXiv ID)**
-  - Example: "Method A [arXiv:2301.12345] achieves X but fails at Y"
-  - Example: "Method B [arXiv:2302.67890] proposes Z but has limitation W"
-- Motivation (why this gap matters)
-- Proposed method (with math formulas)
-- **How this improves on cited papers**
-- Expected advantages
-- Evaluation plan (datasets, baselines from the papers you read)
-- Novelty/Feasibility/Impact scores
+Create `$WORKSPACE/ideas/idea_1.md` through `idea_5.md` using template in `references/idea-template.md`.
 
-**❌ REJECTED if:** No arXiv IDs cited, or ideas not connected to searched literature
+**Key requirements:**
+- Each idea must cite ≥2 papers by arXiv ID
+- Use different strategies (see template): combination, simplification, generalization, constraint relaxation, architecture innovation
 
-### 5.4 Generate Ideas 2-5
-
-For each subsequent idea, explicitly try a **different strategy**:
-
-| Idea | Strategy |
-|------|----------|
-| 1 | Combination - merge techniques from 2+ papers |
-| 2 | Simplification - simplify complex method |
-| 3 | Generalization - extend to new domain/task |
-| 4 | Constraint relaxation - remove limiting assumption |
-| 5 | Architecture innovation - novel model design |
-
-Create `idea_2.md`, `idea_3.md`, `idea_4.md`, `idea_5.md`.
+**❌ REJECTED if:** No arXiv IDs cited, or ideas not grounded in literature
 
 **Output:** `$WORKSPACE/ideas/idea_1.md` through `idea_5.md`
 
@@ -519,36 +395,16 @@ Create `idea_2.md`, `idea_3.md`, `idea_4.md`, `idea_5.md`.
 
 ### 6.1 Evaluate All Ideas
 
-Create evaluation matrix:
-
-```markdown
-# Idea Evaluation
-
-| Idea | Title | Novelty | Feasibility | Impact | Total |
-|------|-------|---------|-------------|--------|-------|
-| 1 | ... | 4 | 3 | 4 | 11 |
-| 2 | ... | 5 | 4 | 5 | 14 |
-| 3 | ... | 3 | 5 | 3 | 11 |
-| 4 | ... | 4 | 4 | 4 | 12 |
-| 5 | ... | 3 | 3 | 4 | 10 |
-
-**Selected: Idea 2**
-
-## Selection Rationale
-[Why this idea is most promising - technical innovation, feasibility, impact]
-```
+Score each idea on Novelty/Feasibility/Impact (1-5). Select highest total.
 
 ### 6.2 Enhance Selected Idea
 
-Take the winning idea and create `$WORKSPACE/ideas/selected_idea.md`:
-
-**Enhancements to add:**
-1. More detailed math formulations (complete loss functions, gradients)
-2. Specific architecture choices (layer sizes, activations)
-3. Hyperparameter recommendations
-4. Implementation roadmap
-5. Potential failure modes and mitigations
-6. Detailed experiment design
+Create `$WORKSPACE/ideas/selected_idea.md` with:
+- Detailed math (loss functions, gradients)
+- Architecture choices
+- Hyperparameters
+- Implementation roadmap
+- Failure modes & mitigations
 
 **Output:** `$WORKSPACE/ideas/selected_idea.md`
 
@@ -556,91 +412,14 @@ Take the winning idea and create `$WORKSPACE/ideas/selected_idea.md`:
 
 ## Step 7: Code Survey - Map Idea to Implementations
 
-This step bridges theory and code. For each **atomic concept** in the selected idea, find corresponding implementations in the reference repos.
+Map each **atomic concept** in the selected idea to code in reference repos.
 
-### 7.1 Extract Atomic Concepts
+See `references/code-mapping.md` for detailed template.
 
-From selected_idea.md, list all concepts needing implementation:
-
-```markdown
-## Atomic Concepts to Implement
-
-1. Multi-head Self-Attention
-2. Graph Message Passing
-3. Energy-based Diffusion
-4. Adaptive Diffusivity Function
-5. ...
-```
-
-### 7.2 Survey Codebases
-
-For each concept:
-
-1. Search repos for relevant code:
-   ```bash
-   grep -r "class.*Attention" $WORKSPACE/repos/
-   grep -r "def forward" $WORKSPACE/repos/
-   ```
-
-2. Read and understand the implementation
-
-3. Document the mapping
-
-### 7.3 Create Implementation Report
-
-Write to `$WORKSPACE/ideas/implementation_report.md`:
-
-```markdown
-# Implementation Report
-
-## Selected Idea Summary
-[One paragraph summary]
-
-## Concept-to-Code Mapping
-
-### Concept 1: Multi-head Self-Attention
-
-**Math Formula:**
-$$
-\text{Attention}(Q,K,V) = \text{softmax}\left(\frac{QK^T}{\sqrt{d_k}}\right)V
-$$
-
-**Reference Implementation:**
-- File: `repos/transformer/attention.py`
-- Class: `MultiHeadAttention`
-- Key code:
-```python
-class MultiHeadAttention(nn.Module):
-    def __init__(self, d_model, n_heads):
-        self.d_k = d_model // n_heads
-        self.W_q = nn.Linear(d_model, d_model)
-        # ...
-
-    def forward(self, x):
-        Q = self.W_q(x)
-        # ...
-```
-
-**Adaptation needed:**
-- [What to modify for our idea]
-
----
-
-### Concept 2: Graph Message Passing
-...
-
----
-
-## Implementation Roadmap
-
-1. [ ] Start with Concept X (foundational)
-2. [ ] Build Concept Y on top
-3. [ ] Integrate with Concept Z
-4. [ ] Add training loop from repo W
-
-## Recommended Starting Point
-[Which repo to fork/use as base]
-```
+**Quick steps:**
+1. Extract atomic concepts from `selected_idea.md`
+2. Search repos: `grep -r "class.*Attention" $WORKSPACE/repos/`
+3. Document mapping to `$WORKSPACE/ideas/implementation_report.md`
 
 **Output:** `$WORKSPACE/ideas/implementation_report.md`
 
@@ -648,53 +427,12 @@ class MultiHeadAttention(nn.Module):
 
 ## Step 8: Final Summary
 
-Create `$WORKSPACE/ideas/summary.md`:
-
-```markdown
-# Research Idea Generation Report
-
-## Task
-- Domain: {domain}
-- Focus: {focus}
-- Date: {date}
-
-## Resources Gathered
-- Papers analyzed: X
-- Repositories cloned: Y
-- Key techniques identified: Z
-
-## Ideas Generated
-1. **[Idea 1 title]** - Score: 11
-2. **[Idea 2 title]** - Score: 14 ⭐ SELECTED
-3. **[Idea 3 title]** - Score: 11
-4. **[Idea 4 title]** - Score: 12
-5. **[Idea 5 title]** - Score: 10
-
-## Selected Idea
-**{Title}**
-
-{One paragraph description}
-
-### Key Innovation
-{What makes this novel}
-
-### Implementation Ready
-- Math formulas: ✓ Complete
-- Code references: ✓ Mapped
-- Evaluation plan: ✓ Defined
-
-## Next Steps
-1. Run `/research-pipeline` with `selected_idea.md` as input
-2. Or manually implement following `implementation_report.md`
-
-## Files Generated
-- `task.json` - Task definition
-- `search_results.md` - Search results
-- `prepare_res.md` - Selected repos
-- `ideas/idea_*.md` - 5 generated ideas
-- `ideas/selected_idea.md` - Enhanced best idea
-- `ideas/implementation_report.md` - Code mapping
-```
+Create `$WORKSPACE/ideas/summary.md` with:
+- Task overview (domain, focus)
+- Resources gathered (papers, repos count)
+- All 5 ideas with scores
+- Selected idea details
+- Next steps: `/research-pipeline` or manual implementation
 
 **Output:** `$WORKSPACE/ideas/summary.md`
 
@@ -717,15 +455,11 @@ Before completing, verify:
 ## Integration with Other Skills
 
 **After idea-generation:**
-```
-/research-pipeline  → Implement the selected idea
-```
+- `/research-pipeline` → Implement the selected idea
 
 **To gather more resources:**
-```
-/arxiv "specific topic"     → Search more papers
-/literature-review          → Deep dive into papers
-```
+- `/literature-survey` → Comprehensive paper collection
+- `/write-review-paper` → Synthesize into review
 
 ---
 

package/skills/idea-generation/references/code-mapping.md

@@ -0,0 +1,51 @@
+# Code Survey & Implementation Mapping
+
+Map selected idea concepts to actual code implementations.
+
+## Step 1: Extract Atomic Concepts
+
+From `selected_idea.md`, list concepts needing implementation.
+
+## Step 2: Survey Codebases
+
+```bash
+grep -r "class.*Attention" $WORKSPACE/repos/
+grep -r "def forward" $WORKSPACE/repos/
+```
+
+## Step 3: Implementation Report
+
+Write to `$WORKSPACE/ideas/implementation_report.md`:
+
+```markdown
+# Implementation Report
+
+## Concept-to-Code Mapping
+
+### Concept 1: [Name]
+
+**Math:**
+$$\text{Formula}$$
+
+**Reference:**
+- File: `repos/<repo>/path/to/file.py`
+- Class: `ClassName`
+
+**Adaptation needed:**
+- [Modifications for our idea]
+
+## Implementation Roadmap
+
+1. [ ] Concept X (foundational)
+2. [ ] Concept Y
+3. [ ] Integration
+
+## Recommended Base Repository
+[repo_name] - [reason]
+```
+
+## Quality Checklist
+
+- [ ] Every concept has code reference
+- [ ] No placeholders
+- [ ] Clear roadmap order

package/skills/idea-generation/references/idea-template.md

@@ -1,217 +1,82 @@
 # Idea Template
 
-Use this template for each idea in `~/.openclaw/workspace/ideas/`.
+Use this template for each generated idea (`$WORKSPACE/ideas/idea_N.md`).
 
----
+## Required Sections
 
-# Idea N: [Short Descriptive Title]
+```markdown
+# Idea N: [Title]
 
-## One-Line Summary
-
-[A single sentence that captures the core innovation. Should be understandable without context.]
-
----
+## One-line Summary
+[Single sentence describing the core innovation]
 
 ## Challenges Addressed
+1. [Challenge 1]
+2. [Challenge 2]
 
-What problems in the current research landscape does this idea solve?
-
-- **[Challenge 1]**: [Brief description of the technical limitation]
-- **[Challenge 2]**: [Brief description of unsolved problem]
-- **[Challenge 3]**: [Brief description of key bottleneck]
-
----
+## Existing Methods & Limitations
 
-## Existing Methods & Their Limitations
+### Method A [arXiv:XXXX.XXXXX]
+- **Approach**: [Brief description]
+- **Limitation**: [Why insufficient]
 
-| Method | Paper/Source | Strength | Weakness This Idea Addresses |
-|--------|--------------|----------|------------------------------|
-| [Method A] | [Citation] | [What it does well] | [Limitation] |
-| [Method B] | [Citation] | [What it does well] | [Limitation] |
-| [Method C] | [Citation] | [What it does well] | [Limitation] |
-
----
+### Method B [arXiv:XXXX.XXXXX]
+- **Approach**: [Brief description]
+- **Limitation**: [Why insufficient]
 
 ## Motivation
-
-### Why is this problem important?
-
-[Explain the significance of solving this problem. Who benefits? What applications are enabled?]
-
-### What gap does this fill?
-
-[Describe the specific research gap this idea addresses. Reference the limitations above.]
-
-### Potential impact
-
-[Quantify if possible: "Could improve X metric by Y%" or "Enables new application Z"]
-
----
+[Why this gap matters, what opportunity exists]
 
 ## Proposed Method
 
-### Core Insight
-
-[2-3 sentences describing the key innovation. What is the "aha" moment?]
+### Core Idea
+[Main innovation in 2-3 sentences]
 
 ### Technical Approach
+[Detailed description]
 
-**Overview:**
-
-[1 paragraph high-level description]
-
-**Step-by-step methodology:**
-
-1. **[Step 1 Name]**: [Description]
-   - Input: [what this step takes]
-   - Output: [what this step produces]
-   - Key operation: [main computation]
-
-2. **[Step 2 Name]**: [Description]
-   - Input: ...
-   - Output: ...
-   - Key operation: ...
-
-3. **[Step 3 Name]**: [Description]
-   - ...
-
-### Mathematical Formulation
-
-**Problem Setup:**
-
-Let $X \in \mathbb{R}^{n \times d}$ denote [description]...
-
-**Core Equations:**
-
-```latex
-% Main loss function
-\mathcal{L} = \mathcal{L}_{task} + \lambda \mathcal{L}_{reg}
-
-% Where task loss is:
-\mathcal{L}_{task} = ...
-
-% And regularization term is:
-\mathcal{L}_{reg} = ...
-```
-
-**Key derivations (if applicable):**
-
-[Show important mathematical steps that justify the approach]
-
-### Architecture / Algorithm
-
-```
-Algorithm: [Name]
-Input: [inputs]
-Output: [outputs]
-
-1. Initialize [parameters]
-2. For each [iteration]:
-   a. Compute [something]
-   b. Update [something]
-3. Return [result]
-```
-
-Or for neural architectures:
-
-```
-[Input] → [Layer 1] → [Layer 2] → ... → [Output]
-         (dim: ...)   (dim: ...)         (dim: ...)
-```
+### Key Equations
+$$
+\mathcal{L} = ...
+$$
 
----
+## How This Improves on Cited Papers
+| Paper | Their Limitation | Our Improvement |
+|-------|------------------|-----------------|
+| [A]   | ...              | ...             |
+| [B]   | ...              | ...             |
 
 ## Expected Advantages
-
-Why should this approach work better than existing methods?
-
-- **[Advantage 1]**: [Explanation with reasoning]
-- **[Advantage 2]**: [Explanation with reasoning]
-- **[Advantage 3]**: [Explanation with reasoning]
-
-**Theoretical justification (if applicable):**
-
-[Brief argument for why this should work]
-
----
-
-## Potential Challenges
-
-What could go wrong? How to mitigate?
-
-| Challenge | Risk Level | Mitigation Strategy |
-|-----------|------------|---------------------|
-| [Challenge 1] | High/Med/Low | [How to address] |
-| [Challenge 2] | High/Med/Low | [How to address] |
-| [Challenge 3] | High/Med/Low | [How to address] |
-
----
+1. [Advantage 1]
+2. [Advantage 2]
 
 ## Evaluation Plan
 
 ### Datasets
-
-| Dataset | Task | Size | Why Chosen |
-|---------|------|------|------------|
-| [Dataset 1] | [Task] | [Size] | [Reason] |
-| [Dataset 2] | [Task] | [Size] | [Reason] |
+- [Dataset 1]: [Why chosen]
 
 ### Baselines
-
-| Method | Paper | Why Compare |
-|--------|-------|-------------|
-| [Baseline 1] | [Citation] | [Reason] |
-| [Baseline 2] | [Citation] | [Reason] |
+- [Method from Paper A]
+- [Method from Paper B]
 
 ### Metrics
+- [Metric 1]: [What it measures]
 
-| Metric | Description | Expected Improvement |
-|--------|-------------|---------------------|
-| [Metric 1] | [What it measures] | [X% over baseline] |
-| [Metric 2] | [What it measures] | [Y% over baseline] |
-
-### Ablation Studies
-
-What components to ablate to understand contribution?
-
-1. [Component 1]: Remove/replace to test [hypothesis]
-2. [Component 2]: Remove/replace to test [hypothesis]
-
----
-
-## Scores
-
-| Criterion | Score (1-5) | Justification |
-|-----------|-------------|---------------|
-| **Novelty** | [X] | [Why this score] |
-| **Feasibility** | [X] | [Why this score] |
-| **Impact** | [X] | [Why this score] |
-| **Total** | [Sum] | |
-
----
-
-## Implementation Notes
-
-### Recommended Libraries
-
-- [Library 1]: For [purpose]
-- [Library 2]: For [purpose]
-
-### Reference Code
-
-- [Repo 1](URL): [What to reference]
-- [Repo 2](URL): [What to reference]
-
-### Estimated Effort
-
-- Model implementation: [X days]
-- Data pipeline: [X days]
-- Training & evaluation: [X days]
-- Total: [X days]
+## Scores (1-5)
+- **Novelty**: X/5 - [Justification]
+- **Feasibility**: X/5 - [Justification]
+- **Impact**: X/5 - [Justification]
+- **Total**: XX/15
+```
 
----
+## Idea Strategies
 
-## Related Ideas
+Generate 5 ideas using different strategies:
 
-- **Idea [M]**: [How it relates - could be combined? alternative approach?]
-- **Future extension**: [What could come next after this idea]
+| Idea | Strategy | Description |
+|------|----------|-------------|
+| 1 | Combination | Merge techniques from 2+ papers |
+| 2 | Simplification | Simplify complex method |
+| 3 | Generalization | Extend to new domain/task |
+| 4 | Constraint Relaxation | Remove limiting assumption |
+| 5 | Architecture Innovation | Novel model design |

package/skills/idea-generation/references/reading-long-papers.md

@@ -0,0 +1,43 @@
+# Reading Long Papers Strategy
+
+For papers >50KB or >15k tokens, use chunked reading.
+
+## Step 1: Structure Scan
+
+```bash
+ls -la $WORKSPACE/papers/{arxiv_id}/
+wc -l $WORKSPACE/papers/{arxiv_id}/*.tex
+```
+
+## Step 2: Chunked Reading
+
+Use Read tool with `offset` and `limit`:
+
+```
+Tool: Read
+Arguments:
+  file_path: "$WORKSPACE/papers/2404.04429/main.tex"
+  offset: 1
+  limit: 500    # First 500 lines
+```
+
+## Priority Sections
+
+| Priority | Section | Why |
+|----------|---------|-----|
+| 1 | Abstract | Core contribution |
+| 2 | Method | Technical details |
+| 3 | Experiments | Results |
+| 4 | Conclusion | Limitations |
+
+## Skip These
+
+- Appendix, Acknowledgments
+- Detailed hyperparameter tables
+
+## Quick Extraction
+
+```bash
+grep -n "\\\\section{" main.tex
+sed -n '/\\begin{abstract}/,/\\end{abstract}/p' main.tex
+```

package/skills/literature-survey/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: literature-survey
-description: "Comprehensive literature survey for a research domain. Automatically searches, filters, clusters, and iterates to discover relevant papers. Use for: exploring a new research area, collecting papers for a topic, building a literature database."
+description: "Comprehensive literature survey (100+ papers). Searches, filters, clusters, and iterates for complete coverage. Use for: exploring new research areas, collecting papers systematically, building literature databases. NOT for: summarizing papers you have (use /write-review-paper), finding a specific paper (use arxiv_search), generating ideas (use /idea-generation)."
 metadata:
   {
     "openclaw":
@@ -14,19 +14,6 @@ metadata:
 
 Comprehensive literature discovery workflow for a research domain. This skill searches broadly, filters by relevance, clusters by direction, and iterates to ensure complete coverage.
 
-**Use this skill when:**
-- Exploring a new research area
-- Collecting all relevant papers for a topic
-- Building a comprehensive literature database
-- Need to understand the landscape of a field
-
-**NOT for:**
-- Summarizing papers you already have (use `/literature-review`)
-- Finding a specific paper (use `arxiv_search` directly)
-- Generating research ideas (use `/idea-generation`)
-
----
-
 ## Architecture: Isolated Sub-agent
 
 This survey runs in an **isolated sub-session** to avoid context pollution. The main session only receives the final report.

package/skills/research-pipeline/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: research-pipeline
-description: "End-to-end research automation: from idea to code implementation with literature review, planning, and iterative refinement. Use arxiv, github_search, and exec tools."
+description: "End-to-end research automation: idea → literature → plan → implement → review → iterate. Use for: implementing a specific research idea, full ML research workflow. NOT for: just exploring literature (use /literature-survey), just generating ideas (use /idea-generation), just writing review (use /write-review-paper)."
 metadata:
   {
     "openclaw":

package/skills/write-review-paper/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: write-review-paper
-description: "Write a literature review or survey paper from EXISTING papers. Guides through reading, note-taking, synthesis, and structured writing. Use after /literature-survey has collected papers."
+description: "Write literature review or survey paper from EXISTING papers. Guides reading strategy, note-taking, synthesis, and academic writing. Use for: writing thesis lit review, standalone survey papers, synthesizing findings. NOT for: searching new papers (use /literature-survey), generating research ideas (use /idea-generation)."
 metadata:
   {
     "openclaw":
@@ -14,18 +14,6 @@ metadata:
 
 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.
 
-**Use this skill when:**
-- You have collected papers (via `/literature-survey` or manually)
-- Need to write a literature review section for a thesis/paper
-- Writing a standalone survey paper
-- Synthesizing findings across multiple papers
-
-**NOT for:**
-- Searching and discovering new papers (use `/literature-survey`)
-- Generating research ideas (use `/idea-generation`)
-
----
-
 ## Prerequisites
 
 Before starting, ensure you have:
@@ -70,62 +58,7 @@ Create `$WORKSPACE/review/reading_plan.md`:
 
 ### 1.2 Reading Notes Template
 
-For each paper, create `$WORKSPACE/review/notes/{paper_id}.md`:
-
-```markdown
-# [Paper Title]
-
-**ID**: [arxiv_id / DOI]
-**Authors**: [author list]
-**Year**: [year]
-**Venue**: [conference/journal]
-**Priority**: P1/P2/P3
-
-## One-sentence Summary
-[用一句话概括这篇论文的核心贡献]
-
-## Problem & Motivation
-- 研究什么问题？
-- 为什么重要？
-- 现有方法的不足？
-
-## Method
-### Core Idea
-[核心创新点]
-
-### Technical Approach
-[关键技术细节]
-
-### Key Equations
-```latex
-[重要公式]
-```
-
-## Experiments
-- **Datasets**: [使用的数据集]
-- **Baselines**: [对比方法]
-- **Metrics**: [评价指标]
-- **Key Results**: [主要结论]
-
-## Strengths
-1. [优点1]
-2. [优点2]
-
-## Limitations
-1. [局限1]
-2. [局限2]
-
-## Connections
-- 与 [paper_x] 的关系：[描述]
-- 改进了 [method_y]：[如何改进]
-- 被 [paper_z] 引用/扩展：[描述]
-
-## Quotes for Citation
-> "[重要原文]" (Section X, Page Y)
-
-## My Comments
-[你的思考、疑问、可能的改进方向]
-```
+For each paper, create `$WORKSPACE/review/notes/{paper_id}.md` using template in `references/note-template.md`.
 
 ---
 
@@ -203,126 +136,19 @@ Create `$WORKSPACE/review/taxonomy.md`:
 
 ### 3.1 Survey Paper Template
 
-Create `$WORKSPACE/review/draft.md`:
-
-```markdown
-# [Survey Title]: A Comprehensive Review
-
-## Abstract
-[Background - 1 sentence]
-[Problem - 1 sentence]
-[What this survey covers - 2 sentences]
-[Key findings - 2 sentences]
-[Conclusion - 1 sentence]
-
-## 1. Introduction
-
-### 1.1 Background and Motivation
-[为什么这个领域重要？]
-[当前研究的热度和趋势]
-
-### 1.2 Scope of This Survey
-[本综述覆盖的范围]
-[与已有综述的区别]
-
-### 1.3 Contributions
-本文的主要贡献：
-1. [贡献1]
-2. [贡献2]
-3. [贡献3]
-
-### 1.4 Organization
-本文结构如下：
-- Section 2: [内容]
-- Section 3: [内容]
-- ...
-
-## 2. Background and Preliminaries
-
-### 2.1 Problem Definition
-[正式定义研究问题]
+Create `$WORKSPACE/review/draft.md` using template in `references/survey-template.md`.
 
-### 2.2 Key Concepts
-[核心概念解释]
-
-### 2.3 Evaluation Metrics
-[常用评价指标]
-
-## 3. Taxonomy of Methods
-
-### 3.1 Category A: [Name]
-
-#### 3.1.1 Subcategory A.1
-[方法描述]
-[代表性工作]
-
-#### 3.1.2 Subcategory A.2
-...
-
-### 3.2 Category B: [Name]
-...
-
-## 4. Comparative Analysis
-
-### 4.1 Quantitative Comparison
-[对比表格]
-[结果分析]
-
-### 4.2 Qualitative Comparison
-[方法特点对比]
-[适用场景分析]
-
-## 5. Datasets and Benchmarks
-
-### 5.1 Public Datasets
-| Dataset | Size | Source | Features |
-|---------|------|--------|----------|
-| ... | ... | ... | ... |
-
-### 5.2 Benchmark Protocols
-[常用的实验设置]
-
-## 6. Challenges and Future Directions
-
-### 6.1 Open Challenges
-1. **Challenge 1**: [描述]
-2. **Challenge 2**: [描述]
-
-### 6.2 Emerging Trends
-1. **Trend 1**: [描述]
-2. **Trend 2**: [描述]
-
-### 6.3 Recommended Research Directions
-1. [方向1]
-2. [方向2]
-
-## 7. Conclusion
-[总结主要发现]
-[对领域的展望]
-
-## References
-[BibTeX entries]
-```
+Key sections: Abstract → Introduction → Background → Taxonomy → Comparison → Datasets → Future Directions → Conclusion
 
 ### 3.2 Thesis Literature Review Template
 
-For a thesis chapter, use this structure:
-
+For a thesis chapter:
 ```markdown
 # Chapter 2: Literature Review
-
 ## 2.1 Introduction
-[本章目标和结构]
-
 ## 2.2 [Topic Area 1]
-[相关工作综述]
-
 ## 2.3 [Topic Area 2]
-[相关工作综述]
-
 ## 2.4 Summary and Research Gaps
-[总结现有工作的不足]
-[引出你的研究问题]
 ```
 
 ---

package/skills/write-review-paper/references/survey-template.md

@@ -0,0 +1,70 @@
+# Survey Paper Template
+
+Use this template for `$WORKSPACE/review/draft.md`:
+
+```markdown
+# [Survey Title]: A Comprehensive Review
+
+## Abstract
+[Background - 1 sentence]
+[Problem - 1 sentence]
+[What this survey covers - 2 sentences]
+[Key findings - 2 sentences]
+[Conclusion - 1 sentence]
+## 1. Introduction
+
+### 1.1 Background and Motivation
+[Why this field matters]
+
+### 1.2 Scope of This Survey
+[Coverage and differentiation from existing surveys]
+
+### 1.3 Contributions
+1. [Contribution 1]
+2. [Contribution 2]
+
+### 1.4 Organization
+- Section 2: [content]
+- Section 3: [content]
+
+## 2. Background and Preliminaries
+
+### 2.1 Problem Definition
+### 2.2 Key Concepts
+### 2.3 Evaluation Metrics
+
+## 3. Taxonomy of Methods
+
+### 3.1 Category A: [Name]
+#### 3.1.1 Subcategory A.1
+[Method description, representative works]
+
+### 3.2 Category B: [Name]
+...
+
+## 4. Comparative Analysis
+
+### 4.1 Quantitative Comparison
+[Tables, results analysis]
+
+### 4.2 Qualitative Comparison
+[Method characteristics, suitable scenarios]
+
+## 5. Datasets and Benchmarks
+
+### 5.1 Public Datasets
+| Dataset | Size | Source | Features |
+|---------|------|--------|----------|
+
+### 5.2 Benchmark Protocols
+
+## 6. Challenges and Future Directions
+
+### 6.1 Open Challenges
+### 6.2 Emerging Trends
+### 6.3 Recommended Research Directions
+
+## 7. Conclusion
+
+## References
+```