maestro-flow 0.5.31 → 0.5.33

package/.agents/skills/scholar-publish/phases/03-promotion.md

@@ -0,0 +1,255 @@
+# Phase 3: Promotion Content Creation
+
+Generate platform-specific promotional content for the accepted paper across Twitter/X, LinkedIn, and blog post formats.
+
+## Objective
+
+- Create a Twitter/X thread with hook, content, and call-to-action
+- Write a LinkedIn post with professional tone and practical implications
+- Draft a blog post for broader audience accessibility
+- Provide multi-platform scheduling strategy
+
+## Input
+
+- `paperContext`: title, venue, contributions, authors
+- `promotionPlatforms`: twitter | linkedin | blog | all
+
+## Execution
+
+### Step 3.1: Extract Promotion Angles
+
+From the paper, identify angles optimized for social engagement:
+
+```
+PROMOTION_ANGLES:
+  hook: [attention-grabbing one-liner about the result]
+  problem_framing: [why this matters to a broad audience]
+  key_result: [most impressive number or finding]
+  practical_impact: [real-world application or implication]
+  novelty: [what makes this different from prior work]
+  visual_highlight: [best figure for social sharing]
+```
+
+### Step 3.2: Twitter/X Thread (if platforms includes twitter)
+
+Create a thread of 5-8 tweets following this structure:
+
+**Tweet 1 - Hook** (max 280 chars):
+- Attention-grabbing summary of the result
+- Use 1-2 relevant emoji sparingly
+- Include a compelling figure if possible
+- Example format: "Our new paper at [Venue]: [Key finding]! [Brief why it matters]"
+
+**Tweet 2 - Problem** (max 280 chars):
+- What problem does the paper solve?
+- Why existing solutions fall short
+- Frame as something the audience has experienced
+
+**Tweet 3 - Method** (max 280 chars):
+- High-level approach in simple terms
+- Avoid jargon - explain as if to a smart non-expert
+- Include architecture figure if helpful
+
+**Tweet 4-5 - Key Results** (max 280 chars each):
+- One major result per tweet
+- Lead with the number or comparison
+- Include result figure or table screenshot
+- Example: "Result: X% improvement over [baseline] on [benchmark]"
+
+**Tweet 6 - Broader Impact** (max 280 chars):
+- Practical implications
+- Who benefits and how
+- Future possibilities
+
+**Tweet 7 - Call to Action** (max 280 chars):
+- Link to paper (arXiv, conference proceedings)
+- Link to code repository (if available)
+- Link to demo (if available)
+- Relevant hashtags (3-5 max): #MachineLearning #NeurIPS2026 etc.
+- Tag co-authors: @author1 @author2
+
+**Thread Guidelines**:
+- Number each tweet (1/N format) for readability
+- Each tweet should make sense standalone
+- First tweet is most important (it appears in timeline)
+- Include 1-2 figures across the thread
+- Post thread at peak engagement times (9-11 AM or 1-3 PM in target timezone)
+
+### Step 3.3: LinkedIn Post (if platforms includes linkedin)
+
+Write a professional post of 3-5 paragraphs:
+
+**Structure**:
+
+**Paragraph 1 - Announcement** (2-3 sentences):
+- Paper accepted announcement
+- Conference name and significance
+- One-line summary of contribution
+
+**Paragraph 2 - Problem and Motivation** (3-4 sentences):
+- The challenge being addressed
+- Why it matters in industry/practice
+- Connect to current trends or real-world scenarios
+
+**Paragraph 3 - Approach and Results** (3-4 sentences):
+- Method overview in accessible terms
+- Key results with metrics
+- Comparison to prior state-of-the-art
+
+**Paragraph 4 - Implications** (2-3 sentences):
+- Practical applications
+- Who can benefit from this work
+- Open-source availability (if applicable)
+
+**Paragraph 5 - Acknowledgments and Links** (2-3 sentences):
+- Thank co-authors and advisors
+- Thank funding sources (if appropriate)
+- Links to paper, code, demo
+
+**Formatting**:
+- Add a key figure or diagram as the post image
+- Use line breaks between paragraphs for readability
+- Add 3-5 relevant hashtags at the end
+- Tag co-authors and institutions
+- Professional but approachable tone
+
+### Step 3.4: Blog Post (if platforms includes blog)
+
+Draft a blog post of 800-1500 words:
+
+**Structure**:
+
+**Title**: Engaging, non-jargon title that captures the key finding
+- Good: "Teaching AI to Understand Complex Tables: Our NeurIPS 2026 Paper"
+- Avoid: "A Novel Multi-Modal Transformer for Semi-Structured Data Parsing"
+
+**Introduction** (150-200 words):
+- Hook the reader with a relatable scenario
+- State the problem in everyday terms
+- Preview the solution and its impact
+
+**The Problem** (150-200 words):
+- Explain the challenge for a general technical audience
+- Use analogies or real-world examples
+- Why existing approaches are insufficient
+
+**Our Approach** (200-300 words):
+- Method overview with intuitive explanations
+- Key innovation described simply
+- Include the main architecture figure with detailed caption
+- Use analogies: "Think of it like..."
+
+**Results** (200-300 words):
+- Key findings with context
+- Include 1-2 figures with explanations
+- Compare to baselines in plain language
+- Highlight practical implications of numbers
+
+**What This Means** (100-150 words):
+- Real-world applications
+- Limitations honestly acknowledged
+- Future research directions
+
+**Try It Yourself** (50-100 words):
+- Links to paper, code, demo
+- How to get started with the code
+- Contact information for collaboration
+
+**Blog Post Guidelines**:
+- Write for a technical but non-specialist audience
+- Avoid unexplained acronyms
+- Include alt text descriptions for all figures
+- Link to related blog posts or resources
+- Add a TL;DR summary at the top or bottom
+
+### Step 3.5: Multi-Platform Scheduling Strategy
+
+Provide timing recommendations:
+
+**Posting Schedule**:
+| Platform | Timing | Rationale |
+|----------|--------|-----------|
+| Twitter/X | 1 week before conference | Build anticipation |
+| LinkedIn | Day of / day after acceptance | Professional announcement |
+| Blog | 1-2 weeks after acceptance | Allows time for polished content |
+
+**Cross-Platform Coordination**:
+- Coordinate posting times with co-authors
+- Share each other's posts for amplification
+- Engage with comments within first 2 hours (algorithm boost)
+- Re-share during the conference week with "presenting today/tomorrow" update
+- Pin the thread/post during conference dates
+
+**Engagement Best Practices**:
+- Respond to all substantive comments
+- Thank people who share your work
+- Follow back researchers who engage
+- Include accessible figure descriptions for screen readers
+- Consider posting in multiple languages if audience is international
+
+### Step 3.6: Generate Output File
+
+Write `promotion-content.md` with the following structure:
+
+```markdown
+# Promotion Content: [Paper Title]
+## Conference: [Venue] | Platforms: [selected platforms]
+
+## Twitter/X Thread
+### Tweet 1/N (Hook)
+[content]
+
+### Tweet 2/N (Problem)
+[content]
+
+[... remaining tweets]
+
+### Thread Figures
+- Figure 1: [description, source from paper]
+- Figure 2: [description, source from paper]
+
+---
+
+## LinkedIn Post
+[full post content]
+
+### Post Image
+- Recommended: [figure description]
+
+### Hashtags
+[hashtag list]
+
+---
+
+## Blog Post
+
+### Title: [blog title]
+
+### TL;DR
+[2-3 sentence summary]
+
+[Full blog post content]
+
+---
+
+## Scheduling Strategy
+[From Step 3.5]
+
+## Engagement Checklist
+- [ ] Co-authors tagged and notified
+- [ ] Figures exported in high resolution for social sharing
+- [ ] All links verified (paper, code, demo)
+- [ ] QR codes generated (if using for conference)
+- [ ] Hashtags researched for current relevance
+- [ ] Accessible descriptions added for figures
+- [ ] Cross-platform posting schedule agreed with co-authors
+```
+
+## Output
+
+- **File**: `promotion-content.md` in working directory
+- **track_tasks**: Mark Phase 3 completed with platforms summary
+
+## Next Phase
+
+Return to orchestrator. This is the final phase - workflow complete after execution.

package/.agents/skills/scholar-rebuttal-pro/README.md

@@ -0,0 +1,313 @@
+# Scholar Rebuttal Pro - 使用指南
+
+## 概述
+
+**scholar-rebuttal-pro** 是一个增强版的学术论文审稿意见回复 skill，整合了 Gemini/CLI 协作分析和多视角讨论机制，用于生成结构化、证据支撑的 rebuttal 文档。
+
+## 核心特性
+
+### 1. 五阶段工作流
+
+```
+Phase 1: 审稿意见解析与分类
+   ↓ (使用 Gemini CLI 语义分析)
+Phase 2: 多视角讨论
+   ↓ (作者/审稿人/专家三方视角)
+Phase 3: 策略制定
+   ↓ (Accept/Defend/Clarify/Experiment)
+Phase 4: Rebuttal 撰写
+   ↓ (会议特定模板 + 专业语气优化)
+Phase 5: 质量验证
+   ↓ (完整性/专业性/说服力评估)
+```
+
+### 2. CLI 协作分析
+
+- **Phase 1**: Gemini CLI 语义分析，自动分类审稿意见
+- **Phase 2**: 多视角讨论（可选 team-ultra-analyze）
+- **Phase 3**: CLI 搜索论文内容，提取支撑证据
+- **Phase 5**: CLI 质量验证，生成改进建议
+
+### 3. 多视角讨论
+
+模拟三方视角进行讨论：
+- **作者视角**: 如何最有效回应
+- **审稿人视角**: 什么样的回复最有说服力
+- **领域专家视角**: 技术准确性和学术规范
+
+### 4. 会议特定策略
+
+支持主流会议模板：
+- **ML Conferences**: NeurIPS/ICML/ICLR
+- **CV Conferences**: CVPR/ECCV/ICCV
+- **NLP Conferences**: ACL/EMNLP
+- **Generic**: 通用模板
+
+## 使用方法
+
+### 基本调用
+
+```bash
+# 方式 1: 直接调用 skill
+scholar-rebuttal-pro
+
+# 方式 2: 通过触发词
+"帮我回复审稿意见"
+"respond to reviewers"
+"rebuttal"
+```
+
+### 输入格式
+
+支持三种输入方式：
+
+1. **文件路径**:
+   ```
+   reviews.txt
+   reviewer-comments.md
+   reviews.pdf
+   ```
+
+2. **内联文本**: 直接粘贴审稿意见
+
+3. **结构化 JSON**: 预解析的审稿结构
+
+### 交互式配置
+
+Skill 启动时会询问三个配置：
+
+#### 1. 自动模式
+- **交互模式 (推荐)**: 每阶段后确认，可调整策略
+- **自动模式**: 一次性执行所有阶段
+
+#### 2. 论文来源
+- **提供路径**: 指定论文 PDF/LaTeX 路径
+- **当前目录**: 自动搜索当前目录
+- **仅审稿意见**: 不分析论文内容
+
+#### 3. 会议类型
+- **ML Conferences**: NeurIPS/ICML/ICLR 策略
+- **CV Conferences**: CVPR/ECCV/ICCV 策略
+- **NLP Conferences**: ACL/EMNLP 策略
+- **Generic**: 通用模板
+
+## 输出产物
+
+### Phase 1 输出
+- `review-analysis.json` - 结构化分类结果
+- `comment-classification.md` - 人类可读的分类报告
+
+### Phase 2 输出
+- `discussion-log.md` - 完整讨论记录
+- `consensus-strategies.json` - 共识策略
+
+### Phase 3 输出
+- `strategy-matrix.md` - 策略矩阵
+- `evidence-references.json` - 证据引用
+
+### Phase 4 输出
+- `rebuttal.md` - 最终 rebuttal 文档
+- `rebuttal-draft-v1-{timestamp}.md` - 版本化草稿
+
+### Phase 5 输出
+- `quality-report.md` - 质量评估报告
+- `improvement-suggestions.json` - 改进建议
+
+所有输出文件位于: `.workflow/.scratchpad/`
+
+## 架构设计
+
+### Orchestrator + Phases 模式
+
+```
+scholar-rebuttal-pro/
+├── SKILL.md                          # 协调器（纯协调逻辑）
+└── phases/
+    ├── 01-review-parsing.md          # Phase 1 完整执行协议
+    ├── 02-multi-perspective-discussion.md  # Phase 2 + Compact Sentinel
+    ├── 03-strategy-formulation.md    # Phase 3 + Compact Sentinel
+    ├── 04-rebuttal-writing.md        # Phase 4 + Compact Sentinel
+    └── 05-quality-validation.md      # Phase 5 完整执行协议
+```
+
+### 渐进式加载
+
+- **SKILL.md**: 仅包含协调逻辑和 `Ref:` 标记
+- **Phase 文件**: 按需加载，包含完整执行细节
+- **Compact Recovery**: Phase 2/3/4 包含 sentinel 机制
+
+### track_tasks 模式
+
+```json
+// Phase 执行中（任务展开）
+[
+  {"content": "Phase 1: Review Parsing", "status": "completed"},
+  {"content": "Phase 2: Multi-Perspective Discussion", "status": "in_progress"},
+  {"content": "  → Author perspective", "status": "in_progress"},
+  {"content": "  → Reviewer perspective", "status": "pending"},
+  {"content": "  → Expert perspective", "status": "pending"},
+  {"content": "Phase 3: Strategy Formulation", "status": "pending"}
+]
+
+// Phase 完成后（任务折叠）
+[
+  {"content": "Phase 1: Review Parsing", "status": "completed"},
+  {"content": "Phase 2: Multi-Perspective Discussion", "status": "completed"},
+  {"content": "Phase 3: Strategy Formulation", "status": "pending"}
+]
+```
+
+## CLI 集成示例
+
+### Phase 1 - 审稿意见分类
+
+```bash
+ccw cli -p "PURPOSE: Parse and classify reviewer comments by type and severity
+TASK: • Parse comment structure • Classify by severity • Extract key concerns
+MODE: analysis
+CONTEXT: @reviews.txt
+EXPECTED: JSON with classification results" \
+--tool gemini --mode analysis --rule analysis-analyze-technical-document
+```
+
+### Phase 3 - 证据搜索
+
+```bash
+ccw cli -p "PURPOSE: Search paper content for evidence supporting response strategies
+TASK: • Locate relevant sections • Extract supporting data • Identify evidence gaps
+MODE: analysis
+CONTEXT: @paper.pdf
+EXPECTED: Evidence map with file:line references" \
+--tool gemini --mode analysis
+```
+
+### Phase 5 - 质量验证
+
+```bash
+ccw cli -p "PURPOSE: Validate rebuttal quality (completeness, professionalism, persuasiveness)
+TASK: • Check all comments addressed • Assess tone • Evaluate evidence strength
+MODE: analysis
+CONTEXT: @rebuttal.md
+EXPECTED: Quality report with improvement suggestions" \
+--tool gemini --mode analysis
+```
+
+## 会议特定策略
+
+### NeurIPS/ICML/ICLR (ML Conferences)
+- 强调概念新颖性和 broader impact
+- 理论严谨性和方法论贡献
+- 实验彻底性和局限性讨论
+
+### CVPR/ECCV/ICCV (CV Conferences)
+- 严格一页限制
+- 禁止外部链接
+- Champion 审稿人策略
+
+### ACL/EMNLP (NLP Conferences)
+- 语言学意义和适当性
+- 伦理考量和数据来源
+- 实用性和可扩展性
+
+### Generic Template
+- 适用于所有会议的通用策略
+- 四大响应策略: Accept/Defend/Clarify/Experiment
+- 五大成功模式（基于 ICLR Spotlight 论文分析）
+
+## 最佳实践
+
+### 1. 准备工作
+- 准备审稿意见文件（txt/md/pdf）
+- 准备论文源文件（PDF/LaTeX）
+- 了解目标会议的特定要求
+
+### 2. 执行流程
+- 使用交互模式（推荐）以便在每阶段后调整
+- 仔细审查 Phase 2 的多视角讨论结果
+- 在 Phase 3 验证证据引用的准确性
+- Phase 4 后人工审查 rebuttal 草稿
+- 根据 Phase 5 的改进建议优化
+
+### 3. 质量保证
+- 确保所有审稿意见都有回应
+- 每个回应都有充分的证据支撑
+- 保持专业、尊重的语气
+- 避免过度承诺（只承诺可行的改进）
+
+### 4. 会议特定注意事项
+- **NeurIPS**: 强调 broader impact 和伦理考量
+- **ICML**: 突出理论贡献和数学证明
+- **ICLR**: 详细讨论局限性，披露 LLM 使用
+- **CVPR**: 严格遵守一页限制，不使用外部链接
+- **ACL**: 强调语言学意义和伦理声明
+
+## 故障排除
+
+### 问题 1: CLI 执行失败
+**解决方案**: 检查 `~/.claude/cli-tools.json` 配置，确保 gemini 工具已启用
+
+### 问题 2: 论文路径无效
+**解决方案**: 使用"仅审稿意见"模式，手动提供证据
+
+### 问题 3: 会议模板未找到
+**解决方案**: 使用 Generic 模板，手动调整会议特定要求
+
+### 问题 4: Phase 执行中断
+**解决方案**: Skill 支持断点恢复，重新调用会从中断处继续
+
+## 扩展性
+
+### 添加新会议模板
+
+1. 在 `G:\github_lib\claude-scholar\skills\review-response\references\` 添加模板
+2. 或在 `d:\ccws\.workflow\参考文档1\` 添加自定义模板
+3. Phase 4 会自动搜索并加载
+
+### 自定义响应策略
+
+编辑 Phase 3 的策略映射规则，添加新的响应类型。
+
+### 集成其他 CLI 工具
+
+修改 Phase 1/3/5 的 CLI 调用，替换 `--tool gemini` 为其他工具（qwen/codex）。
+
+## 相关 Skills
+
+- **scholar-review**: 论文自审和审稿流程
+- **scholar-writing**: 端到端论文撰写
+- **scholar-publish**: 会后准备（演讲/海报）
+
+## 技术细节
+
+### Compact Recovery 机制
+
+Phase 2/3/4 包含 compact sentinel，防止长上下文压缩导致协议丢失：
+
+```markdown
+> **📌 COMPACT SENTINEL [Phase 2: multi-perspective-discussion]**
+> This phase contains 4 execution steps (Step 2.1 — 2.4).
+> If you can read this sentinel but cannot find the full Step protocol below, context has been compressed.
+> Recovery: `read_file("phases/02-multi-perspective-discussion.md")`
+```
+
+### Checkpoint 验证
+
+关键执行步骤前包含 checkpoint：
+
+```markdown
+> **⚠️ CHECKPOINT**: Before proceeding, verify:
+> 1. This phase is track_tasks `in_progress` (active phase protection)
+> 2. Full protocol (Step N.X — N.M) is in active memory, not just sentinel
+> If only sentinel remains → `read_file("phases/0N-xxx.md")` now.
+```
+
+## 版本信息
+
+- **Skill Version**: 1.0.0
+- **Created**: 2026-03-02
+- **Based on**: claude-scholar review-response skill
+- **Enhancements**: CLI integration, multi-perspective discussion, conference templates
+
+## 贡献与反馈
+
+如需改进或报告问题，请联系项目维护者。