@wentorai/research-plugins 1.2.2 → 1.3.0

package/skills/writing/composition/opendraft-thesis-guide/SKILL.md

@@ -1,200 +0,0 @@
----
-name: opendraft-thesis-guide
-description: "AI thesis writer with specialized agents and verified citations"
-metadata:
-  openclaw:
-    emoji: "📖"
-    category: "writing"
-    subcategory: "composition"
-    keywords: ["thesis writing", "AI writer", "multi-agent", "dissertation", "academic writing", "verified citations"]
-    source: "https://github.com/federicodeponte/opendraft"
----
-
-# OpenDraft Thesis Writing Guide
-
-## Overview
-
-OpenDraft is an AI-powered thesis and dissertation writing assistant that uses 19 specialized agents to produce long-form academic documents with verified citations. Each agent handles a specific aspect — outline generation, literature integration, argument construction, methodology writing, and citation verification. Produces coherent multi-chapter documents while ensuring all references are real and accessible.
-
-## Agent Architecture
-
-```
-User Input (topic, outline, notes)
-         ↓
-  Orchestrator Agent
-  ├── Outline Agent (structure chapters)
-  ├── Literature Agent (find relevant papers)
-  ├── Introduction Agent (background + motivation)
-  ├── Theory Agent (theoretical framework)
-  ├── Methodology Agent (research design)
-  ├── Analysis Agent (results + discussion)
-  ├── Conclusion Agent (summary + future work)
-  ├── Citation Agent (verify all references)
-  ├── Consistency Agent (terminology + notation)
-  ├── Style Agent (academic tone + flow)
-  └── Formatting Agent (LaTeX/Markdown output)
-         ↓
-  Verified Thesis Document
-```
-
-## Usage
-
-```python
-from opendraft import ThesisWriter
-
-writer = ThesisWriter(
-    llm_provider="anthropic",
-    citation_backend="semantic_scholar",
-)
-
-# Generate thesis from topic and outline
-thesis = writer.write(
-    title="Efficient Attention Mechanisms for Long-Document "
-          "Understanding",
-    outline={
-        "chapters": [
-            {"title": "Introduction",
-             "points": ["Motivation", "Research questions",
-                        "Contributions"]},
-            {"title": "Background",
-             "points": ["Transformer architecture",
-                        "Attention complexity",
-                        "Existing solutions"]},
-            {"title": "Methodology",
-             "points": ["Proposed approach",
-                        "Architecture design",
-                        "Training procedure"]},
-            {"title": "Experiments",
-             "points": ["Datasets", "Baselines", "Results"]},
-            {"title": "Discussion",
-             "points": ["Analysis", "Limitations",
-                        "Future work"]},
-            {"title": "Conclusion",
-             "points": ["Summary", "Implications"]},
-        ],
-    },
-    notes_dir="./research_notes/",
-    bibliography="references.bib",
-    target_pages=80,
-)
-
-thesis.save("thesis.tex")
-thesis.save("thesis.md")
-```
-
-## Citation Verification
-
-```python
-# All citations are verified against real databases
-verification = thesis.citation_report()
-
-print(f"Total citations: {verification.total}")
-print(f"Verified (DOI): {verification.verified_doi}")
-print(f"Verified (title match): {verification.verified_title}")
-print(f"Unverified: {verification.unverified}")
-
-# List unverified citations
-for cite in verification.unverified_list:
-    print(f"  WARNING: {cite.key} — {cite.reason}")
-
-# All verified references have real DOIs
-for cite in verification.verified_list[:5]:
-    print(f"  OK: {cite.key} — DOI: {cite.doi}")
-```
-
-## Chapter-by-Chapter Writing
-
-```python
-# Write individual chapters with full context
-chapter = writer.write_chapter(
-    chapter_title="Methodology",
-    preceding_chapters=["introduction.tex", "background.tex"],
-    key_points=[
-        "Describe the sparse attention pattern learning algorithm",
-        "Formal complexity analysis: O(n * sqrt(n))",
-        "Training with progressive sequence length curriculum",
-    ],
-    target_length=5000,  # words
-    include_figures=True,
-)
-
-print(chapter.text[:500])
-print(f"Citations used: {len(chapter.citations)}")
-print(f"Figures generated: {len(chapter.figures)}")
-```
-
-## Literature Integration
-
-```python
-# Automatically weave citations into text
-section = writer.write_with_literature(
-    topic="linear attention mechanisms",
-    style="survey",    # survey, argumentative, analytical
-    max_papers=20,
-    recency_bias=0.7,  # Prefer recent papers
-)
-
-# Literature agent:
-# 1. Searches Semantic Scholar for relevant papers
-# 2. Reads abstracts and key findings
-# 3. Groups papers by approach/contribution
-# 4. Weaves into coherent narrative with citations
-```
-
-## Revision and Editing
-
-```python
-# Iterative revision
-revised = writer.revise(
-    manuscript="thesis.tex",
-    feedback=[
-        "Strengthen the motivation in Chapter 1",
-        "Add more quantitative comparisons in Chapter 4",
-        "The notation in Chapter 3 is inconsistent with Chapter 2",
-    ],
-)
-
-# Style check
-style_issues = writer.check_style("thesis.tex")
-for issue in style_issues:
-    print(f"[{issue.type}] Ch.{issue.chapter}, "
-          f"p.{issue.paragraph}: {issue.suggestion}")
-```
-
-## Configuration
-
-```python
-writer = ThesisWriter(
-    llm_provider="anthropic",
-    model="claude-sonnet-4-20250514",
-    writing_config={
-        "person": "first_plural",    # "we"
-        "tense": "present",
-        "formality": "high",
-        "hedging": "moderate",
-    },
-    citation_config={
-        "style": "natbib",
-        "verify_all": True,
-        "min_citations_per_chapter": 10,
-    },
-    output_config={
-        "format": "latex",
-        "template": "university_thesis",
-        "font": "times",
-        "line_spacing": 1.5,
-    },
-)
-```
-
-## Use Cases
-
-1. **Thesis drafting**: Generate first drafts of thesis chapters
-2. **Literature review**: AI-assisted literature review sections
-3. **Revision assistance**: Iterative feedback-based revision
-4. **Citation management**: Verified, real citations throughout
-5. **Template generation**: Structured academic document scaffolding
-
-## References
-
-- [OpenDraft GitHub](https://github.com/federicodeponte/opendraft)

package/skills/writing/composition/paper-debugger-guide/SKILL.md

@@ -1,143 +0,0 @@
----
-name: paper-debugger-guide
-description: "Multi-agent system for academic writing review and improvement"
-version: 1.0.0
-author: wentor-community
-source: https://github.com/PaperDebugger/paperdebugger
-metadata:
-  openclaw:
-    category: "writing"
-    subcategory: "composition"
-    keywords:
-      - paper-review
-      - academic-writing
-      - multi-agent
-      - writing-feedback
-      - peer-review
-      - manuscript-revision
----
-
-# Paper Debugger Guide
-
-A skill for using a plugin-based multi-agent system to review, debug, and improve academic manuscripts before submission. Based on PaperDebugger (1K stars), this skill provides structured workflows for catching logical errors, methodological weaknesses, presentation issues, and formatting problems in research papers.
-
-## Overview
-
-Just as software developers debug code before shipping, researchers should systematically debug their papers before submission. PaperDebugger applies this philosophy by deploying multiple specialized review agents, each focused on a different dimension of paper quality: logical consistency, methodological rigor, clarity of presentation, statistical validity, and formatting compliance.
-
-The result is a pre-submission review process that catches issues human reviewers commonly flag, giving researchers the opportunity to address them proactively and increase their chances of acceptance.
-
-## Review Agent Roles
-
-The system deploys these specialized agents:
-
-**Logic Agent**
-- Checks that conclusions follow from presented evidence
-- Identifies unsupported claims and logical gaps
-- Verifies that the abstract accurately reflects the paper's content
-- Ensures consistency between the introduction's promises and the discussion's delivery
-- Flags circular reasoning and tautological statements
-
-**Methods Agent**
-- Evaluates whether the methodology is appropriate for the research questions
-- Checks for missing methodological details needed for replication
-- Identifies potential confounding variables not addressed
-- Verifies that the sample size and study design support the intended analyses
-- Assesses whether limitations are adequately acknowledged
-
-**Clarity Agent**
-- Identifies sentences that are overly complex or ambiguous
-- Flags jargon used without definition on first occurrence
-- Checks paragraph structure for topic sentences and logical flow
-- Identifies redundant content that could be consolidated
-- Evaluates whether figures and tables are self-explanatory with captions
-
-**Statistics Agent**
-- Verifies that reported statistics match the described analyses
-- Checks for common errors (wrong degrees of freedom, inconsistent n values)
-- Flags p-values reported without effect sizes or confidence intervals
-- Identifies multiple comparison issues without appropriate corrections
-- Validates that data visualization accurately represents the underlying data
-
-**Format Agent**
-- Checks compliance with target journal formatting guidelines
-- Verifies reference formatting consistency
-- Validates figure and table numbering and cross-references
-- Checks adherence to word count limits per section
-- Ensures required sections are present (data availability, conflict of interest, etc.)
-
-## Review Workflow
-
-The complete debugging workflow proceeds through these stages:
-
-**Stage 1: Structural Scan**
-- Parse the manuscript into its constituent sections
-- Verify all expected sections are present and properly ordered
-- Check section lengths against typical proportions and journal requirements
-- Identify orphaned references (cited but not in bibliography, or vice versa)
-- Flag missing elements (acknowledgments, supplementary references, appendices)
-
-**Stage 2: Deep Review**
-- Each specialized agent reviews the full manuscript independently
-- Agents generate issue reports with severity levels (critical, major, minor, suggestion)
-- Issues include specific locations (section, paragraph, sentence) and explanations
-- Each issue includes a suggested fix or improvement direction
-- Agents flag areas outside their expertise for human judgment
-
-**Stage 3: Cross-Agent Synthesis**
-- Merge findings from all agents into a unified report
-- Resolve conflicting recommendations between agents
-- Prioritize issues by severity and effort to address
-- Group related issues that can be addressed together
-- Generate a revision checklist ordered by priority
-
-**Stage 4: Revision Support**
-- For each identified issue, provide specific revision suggestions
-- Generate alternative phrasings for clarity issues
-- Suggest additional analyses or robustness checks for methodology issues
-- Draft missing sections (limitations, future work) based on paper content
-- Help restructure paragraphs or sections when flow issues are identified
-
-## Common Issue Categories
-
-Based on analysis of reviewer feedback patterns, these are the most frequently caught issues:
-
-**High-Priority Issues**
-- Overclaiming: conclusions stronger than the evidence supports
-- Missing baselines: experimental comparisons without appropriate controls
-- Selective reporting: presenting only favorable results
-- Methodology gaps: insufficient detail for replication
-- Statistical errors: incorrect test selection or reporting
-
-**Medium-Priority Issues**
-- Introduction-conclusion misalignment
-- Inconsistent terminology throughout the paper
-- Figures that duplicate information already in tables
-- Literature review missing key recent papers in the field
-- Vague contribution statements
-
-**Low-Priority Issues**
-- Minor grammatical errors
-- Inconsistent formatting of numbers and units
-- Reference style inconsistencies
-- Caption formatting variations
-- Whitespace and layout irregularities
-
-## Integration with Research-Claw
-
-This skill connects with the broader Research-Claw writing workflow:
-
-- Run paper debugging after drafting with composition skills
-- Feed identified methodology issues to the analysis skill for verification
-- Use literature search skills to find missing references
-- Connect with LaTeX skills for formatting compliance fixes
-- Store review reports alongside manuscript versions for revision tracking
-
-## Best Practices
-
-- Run the full debugging suite at least one week before the submission deadline
-- Address critical and major issues first, then iterate on minor issues
-- Have a human co-author review the agent-generated suggestions before implementing
-- Use the debugging report as a conversation starter with collaborators
-- Re-run the suite after major revisions to check for newly introduced issues
-- Keep a log of common issues across papers to identify personal writing patterns

package/skills/writing/composition/paperforge-guide/SKILL.md

@@ -1,205 +0,0 @@
----
-name: paperforge-guide
-description: "AI-assisted academic paper drafting and structuring tool"
-metadata:
-  openclaw:
-    emoji: "🔨"
-    category: "writing"
-    subcategory: "composition"
-    keywords: ["paper writing", "academic drafting", "paper generation", "manuscript", "structured writing", "AI writing"]
-    source: "https://github.com/paperforge/paperforge"
----
-
-# PaperForge Guide
-
-## Overview
-
-PaperForge is an AI-assisted academic paper drafting tool that helps researchers structure, draft, and refine manuscripts. It generates section-by-section drafts from outlines and notes, ensures consistent academic tone, manages citation integration, and provides iterative refinement. Designed to accelerate the writing process while keeping researchers in control of content and argumentation.
-
-## Writing Pipeline
-
-```
-Research Notes + Outline
-         ↓
-   Structure Planning (sections, flow, arguments)
-         ↓
-   Section Drafting (introduction → conclusion)
-         ↓
-   Citation Integration (inline refs from .bib)
-         ↓
-   Consistency Check (terminology, notation, tone)
-         ↓
-   Revision Suggestions (clarity, flow, gaps)
-         ↓
-   Final Manuscript
-```
-
-## Usage
-
-```python
-from paperforge import PaperForge
-
-forge = PaperForge(llm_provider="anthropic")
-
-# Draft from outline and notes
-manuscript = forge.draft(
-    title="Efficient Attention Mechanisms for Long Documents",
-    outline={
-        "introduction": "Transformers limited by quadratic attention. "
-                        "Need sub-quadratic alternatives for long docs.",
-        "related_work": "Linear attention, sparse attention, "
-                        "low-rank approximations.",
-        "method": "Our approach: learned sparse patterns + "
-                  "sliding window hybrid.",
-        "experiments": "Long Range Arena benchmark + document "
-                       "classification tasks.",
-        "conclusion": "2x speedup, comparable accuracy.",
-    },
-    notes_dir="./research_notes/",
-    bibliography="references.bib",
-)
-
-manuscript.save("draft.tex")
-```
-
-## Section-by-Section Drafting
-
-```python
-# Draft individual sections with context
-intro = forge.draft_section(
-    section="introduction",
-    key_points=[
-        "Transformer attention is O(n^2) in sequence length",
-        "Long documents (>4096 tokens) are common in practice",
-        "Existing solutions trade quality for efficiency",
-        "Our contribution: learned sparse patterns",
-    ],
-    style="narrative",      # narrative, technical, concise
-    target_length=800,      # approximate word count
-    bibliography="refs.bib",
-)
-
-print(intro.text)
-print(f"Citations used: {intro.citations}")
-```
-
-## Outline Generation
-
-```python
-# Generate outline from abstract or notes
-outline = forge.generate_outline(
-    abstract="We propose a novel attention mechanism that combines "
-             "learned sparse patterns with sliding window attention "
-             "to achieve sub-quadratic complexity while maintaining "
-             "accuracy on long document tasks.",
-    paper_type="conference",  # conference, journal, workshop
-    target_venue="NeurIPS",
-)
-
-for section in outline.sections:
-    print(f"\n## {section.title}")
-    for point in section.key_points:
-        print(f"  - {point}")
-    print(f"  Target length: {section.target_words} words")
-```
-
-## Citation Integration
-
-```python
-# Automatic citation placement
-section_with_cites = forge.integrate_citations(
-    text=draft_text,
-    bibliography="references.bib",
-    style="authoryear",  # or "numeric"
-    strategy="relevant",  # place citations where claims need support
-)
-
-# Citation suggestions for unsupported claims
-suggestions = forge.suggest_citations(
-    text=draft_text,
-    bibliography="references.bib",
-)
-for s in suggestions:
-    print(f"Claim: '{s.claim[:60]}...'")
-    print(f"  Suggested: {s.suggested_refs}")
-```
-
-## Revision and Refinement
-
-```python
-# Get revision suggestions
-revisions = forge.review(
-    manuscript="draft.tex",
-    checks=[
-        "clarity",           # Unclear sentences
-        "consistency",       # Terminology and notation
-        "flow",              # Logical transitions
-        "redundancy",        # Repeated content
-        "claim_support",     # Unsupported claims
-        "passive_voice",     # Overuse of passive
-    ],
-)
-
-for rev in revisions:
-    print(f"[{rev.category}] Line {rev.line}: {rev.suggestion}")
-    print(f"  Original: {rev.original}")
-    print(f"  Suggested: {rev.revised}")
-
-# Apply selected revisions
-manuscript = forge.apply_revisions(
-    "draft.tex",
-    revisions=[r for r in revisions if r.category == "clarity"],
-)
-```
-
-## Template Support
-
-```python
-# Write to specific venue templates
-manuscript = forge.draft(
-    outline=outline,
-    template="neurips2025",  # or "icml", "acl", "ieee", "arxiv"
-    formatting={
-        "max_pages": 9,       # Excluding references
-        "abstract_words": 200,
-        "font": "times",
-    },
-)
-
-# Available templates
-templates = forge.list_templates()
-# ["neurips2025", "icml2025", "acl2025", "ieee_conference",
-#  "arxiv_preprint", "elegantpaper", "springer_lncs"]
-```
-
-## Configuration
-
-```python
-forge = PaperForge(
-    llm_provider="anthropic",
-    model="claude-sonnet-4-20250514",
-    writing_config={
-        "tone": "formal_academic",
-        "person": "first_plural",    # "we propose..."
-        "tense": "present",          # for methods/results
-        "hedging": "moderate",       # "may", "suggests", "appears"
-    },
-    citation_config={
-        "style": "natbib",
-        "max_refs_per_claim": 3,
-    },
-)
-```
-
-## Use Cases
-
-1. **First drafts**: Generate initial manuscript from structured notes
-2. **Section expansion**: Flesh out outline into full sections
-3. **Revision assistance**: Systematic review for clarity and consistency
-4. **Template conversion**: Reformat between venue styles
-5. **Citation management**: Automated placement and gap detection
-
-## References
-
-- [PaperForge GitHub](https://github.com/paperforge/paperforge)
-- [LaTeX Templates](https://www.latextemplates.com/)