GSDAW Planning Agent

Purpose

Create a coarse-grained writing roadmap based on the Research Brief, Literature Summary, and Methodology Design. This skill operates after aw-discuss-2 confirms consistency between the research and methodology outputs. It breaks the paper into chapter-level phases (not paragraph-level tasks).

When to Trigger

• aw-discuss-2 confirms consistency between Research and Methodology
• User explicitly runs /aw-plan command
• Orchestrator calls this skill during GSDAW pipeline

Workflow

aw-discuss-2 (consistency confirmed)
    │
    ▼
aw-planner  ◄── You are here
    │
    ▼
┌─────────────────────────────────────────────────┐
│  READ INPUT DOCUMENTS                            │
│  1. .planning/research-brief.json (title, constraints, deadline) │
│  2. .planning/literature.md (Related Work structure)            │
│  3. .planning/methodology.md (Method/Experiment/Results chapters) │
│  4. .planning/discuss-2-summary.md (optional consensus notes)    │
└─────────────────────┬───────────────────────────┘
                      │
                      ▼
┌─────────────────────────────────────────────────┐
│  PHASE DECOMPOSITION                             │
│  - Map IMRAD chapters → 6 phases               │
│  - Extract tasks from literature.md + methodology.md │
│  - Calculate estimated duration from deadline    │
└─────────────────────┬───────────────────────────┘
                      │
                      ▼
┌─────────────────────────────────────────────────┐
│  OUTPUT GENERATION                              │
│  1. Write .planning/ROADMAP.md                 │
│  2. Write .planning/STATE.md (initial state)    │
└─────────────────────┬───────────────────────────┘
                      │
                      ▼
Report completion to user

---

Step 1: Read Input Documents

1.1 Read Research Brief

Read .planning/research-brief.json and extract:

Field	Path	Use
Title	title	Paper title in ROADMAP header
Target Venue	constraints.targetVenue	Journal/conference
Deadline	constraints.deadline	Calculate duration
Problem	researchQuestion.problem	Introduction context
Novelty	researchQuestion.novelty	Introduction contributions
Strategy	researchApproach.strategy	Methodology overview

1.2 Read Literature Summary

Read .planning/literature.md and extract:

Field	Use
Categories of related work	Structure Related Work chapter
Key references (must-cite)	Ensure proper citations
Research gaps	Position the contribution
Papers analyzed count	Estimate Related Work scope

1.3 Read Methodology

Read .planning/methodology.md and extract:

Field	Use
Architecture components	Methodology chapter tasks
Datasets	Experiment chapter
Baselines	Experiment chapter
Evaluation metrics	Results chapter
Ablation studies	Results chapter
Figures needed	Coordinate with figure skill
Risks	Discussion chapter

1.4 Read Discuss-2 Summary (if exists)

Read .planning/discuss-2-summary.md for any consensus notes or adjustments from the consistency check.

---

Step 2: Phase Decomposition

Standard IMRAD Phases

Phase	Chapter	Goal
1	Introduction	Establish context and motivate the work
2	Related Work	Position in existing literature
3	Methodology	Describe proposed approach
4	Experiment	Present experimental setup and results
5	Discussion	Interpret results and implications
6	Conclusion	Summarize contributions and future work

Phase Dependencies

Phase 1 (Introduction) → None
Phase 2 (Related Work) → Phase 1 + Literature Summary
Phase 3 (Methodology) → Phase 2 + Methodology Design
Phase 4 (Experiment) → Phase 3
Phase 5 (Discussion) → Phase 4 + Literature Summary
Phase 6 (Conclusion) → Phase 5

Duration Estimation

Calculate estimated total duration based on deadline:

1. Count days from today to deadline
2. Allocate time per phase:

Phase	Default Allocation
Introduction	10% of total
Related Work	15% of total
Methodology	20% of total
Experiment	25% of total
Discussion	20% of total
Conclusion	10% of total

Example: 5 weeks (35 days) until deadline:

• Introduction: 3-4 days
• Related Work: 5 days
• Methodology: 7 days
• Experiment: 9 days
• Discussion: 7 days
• Conclusion: 3-4 days

---

Step 3: Generate ROADMAP.md

Write to .planning/ROADMAP.md:

# GSDAW Roadmap

**Paper:** [title from brief]
**Target:** [journal/conference from brief]
**Estimated Duration:** [X weeks based on deadline]
**Created:** [ISO date]

## Phase Dependencies

Phase 1 (Introduction) Phase 2 (Related Work) ← needs Phase 1 + Literature Phase 3 (Methodology) ← needs Phase 2 + Methodology Phase 4 (Experiment) ← needs Phase 3 Phase 5 (Discussion) ← needs Phase 4 + Literature Phase 6 (Conclusion) ← needs Phase 5

---

## Phase 1: Introduction

**Goal:** Establish research context and motivate the work

**Dependencies:** None

**Estimated Duration:** [X days]

**Tasks:**
- [ ] 1.1 Research background (2-3 paragraphs on problem domain)
- [ ] 1.2 Problem definition (1 paragraph)
- [ ] 1.3 Main contributions (3-4 bullet points, numbered)
- [ ] 1.4 Paper structure overview (brief paragraph)

**Success Criteria:**
1. Reader understands why this problem matters
2. Reader knows the main contributions (3 distinct items)
3. Structure roadmap is clear

---

## Phase 2: Related Work

**Goal:** Position work in context of existing research

**Dependencies:** Phase 1, Literature Summary (from .planning/literature.md)

**Estimated Duration:** [X days]

**Tasks:**
- [ ] 2.1 Categorize existing methods (group by approach type)
- [ ] 2.2 Discuss each category (strengths + weaknesses)
- [ ] 2.3 Highlight research gap (from literature.md)
- [ ] 2.4 Natural transition to proposed method

**Success Criteria:**
1. Covers major related work (from literature.md)
2. Categorization is logical and defensible
3. Gap clearly sets up proposed method

---

## Phase 3: Methodology

**Goal:** Describe the proposed approach in technical detail

**Dependencies:** Phase 2, Methodology Design (from .planning/methodology.md)

**Estimated Duration:** [X days]

**Tasks:**
- [ ] 3.1 Method overview (2-3 paragraphs)
- [ ] 3.2 Architecture components (describe each component)
- [ ] 3.3 Key innovations (explain novelty points)
- [ ] 3.4 Experimental setup overview (datasets, baselines preview)

**Success Criteria:**
1. Technical description is complete and reproducible
2. Reader understands how the method works
3. Innovations are clearly stated and justified

---

## Phase 4: Experiment

**Goal:** Present experimental setup and results

**Dependencies:** Phase 3

**Estimated Duration:** [X days]

**Tasks:**
- [ ] 4.1 Dataset description (statistics, splits, preprocessing)
- [ ] 4.2 Baseline methods (describe each baseline)
- [ ] 4.3 Evaluation metrics (define each metric)
- [ ] 4.4 Main results (compare with baselines)
- [ ] 4.5 Ablation studies (component-wise analysis)
- [ ] 4.6 Statistical significance (if applicable)

**Success Criteria:**
1. All baselines from methodology.md are evaluated
2. Results are reproducible from description
3. Ablation study validates component contributions

---

## Phase 5: Discussion

**Goal:** Interpret results and discuss implications

**Dependencies:** Phase 4, Literature Summary

**Estimated Duration:** [X days]

**Tasks:**
- [ ] 5.1 Result interpretation (what do the numbers mean)
- [ ] 5.2 Comparison with literature (how do results compare to prior work)
- [ ] 5.3 Limitations (honest assessment of weaknesses)
- [ ] 5.4 Risks from methodology.md (address each risk)
- [ ] 5.5 Practical implications (what this means for practitioners)

**Success Criteria:**
1. Results are interpreted honestly (not over-sold)
2. Limitations are acknowledged
3. Comparison with literature is fair and complete

---

## Phase 6: Conclusion

**Goal:** Summarize contributions and future work

**Dependencies:** Phase 5

**Estimated Duration:** [X days]

**Tasks:**
- [ ] 6.1 Summary of contributions (restate main contributions)
- [ ] 6.2 Key findings (1-2 paragraphs on most important results)
- [ ] 6.3 Future work directions
- [ ] 6.4 Closing statement

**Success Criteria:**
1. Contributions are clearly restated
2. Paper ends on a strong note
3. Future work is plausible and interesting

---

## Key Milestones

| Milestone | Target Week | Phase | Deliverable |
|-----------|-------------|-------|-------------|
| Outline Complete | Week 1 | Phases 1-2 | Full paper outline |
| Methodology Done | Week 2 | Phase 3 | Technical description |
| Results Done | Week 3 | Phase 4 | Experiment chapter |
| Discussion Done | Week 4 | Phase 5 | Interpretation |
| Submission Ready | Week 5 | Phase 6 | Final PDF |

---

## Appendix: Input Document Summary

### From research-brief.json

- **Title:** [title]
- **Target Venue:** [venue]
- **Deadline:** [deadline or "not specified"]
- **Problem:** [problem statement]
- **Novelty:** [novelty statement]

### From literature.md

- **Papers Analyzed:** [count]
- **Categories:** [list from literature.md]
- **Key Gaps:** [from literature.md]

### From methodology.md

- **Datasets:** [list]
- **Baselines:** [list]
- **Metrics:** [list]
- **Components:** [list]
- **Figures Needed:** [count]

---

Step 4: Generate STATE.md

Write to .planning/STATE.md:

# GSDAW State

**Paper:** [title]
**Updated:** [ISO date]

## Overall Status

| Phase | Status | Completion |
|-------|--------|------------|
| 1. Introduction | pending | 0% |
| 2. Related Work | pending | 0% |
| 3. Methodology | pending | 0% |
| 4. Experiment | pending | 0% |
| 5. Discussion | pending | 0% |
| 6. Conclusion | pending | 0% |

## Current Phase

None — planning complete, awaiting execution.

## Open Issues

(None yet)

## Notes

(Add notes as work progresses)

---

Step 5: Post-Completion Report

After generating ROADMAP.md and STATE.md, report to user:

写作计划已生成。

论文分为 6 个阶段：
1. Introduction — 建立研究背景
2. Related Work — 相关工作调研
3. Methodology — 方法论
4. Experiment — 实验
5. Discussion — 讨论
6. Conclusion — 结论

下一步：开始 Phase 1 — Introduction

---

Edge Cases

Missing Input Documents

Scenario	Handling
research-brief.json not found	Abort: "错误：未找到研究简报。请先运行 /aw-init。"
literature.md not found	Abort: "错误：未找到文献总结。请先运行 /aw-research。"
methodology.md not found	Abort: "错误：未找到方法论设计。请先运行 /aw-methodology。"
discuss-2-summary.md not found	Skip — proceed without consensus notes

Missing Fields in research-brief.json

Field	Handling
title is null	Use "未命名论文" in ROADMAP
deadline is null	Use "未指定" — don't calculate duration
targetVenue is null	Use "未指定" in ROADMAP

Deadline Too Short

If days until deadline < 21 (3 weeks):

警告：距离截稿日期仅 [X] 天，时间非常紧张。
将优先保证核心章节（Introduction, Methodology, Experiment, Conclusion），
Related Work 和 Discussion 可能需要压缩。

---

Integration Points

Input	Source	Description
Research Brief	.planning/research-brief.json	User-approved brief from Questioner
Literature Summary	.planning/literature.md	From Research Agent
Methodology Design	.planning/methodology.md	From Methodology Agent
Discuss-2 Notes	.planning/discuss-2-summary.md	Consensus from consistency check (optional)

Output	Destination	Consumed By
Roadmap	.planning/ROADMAP.md	User, orchestrator
State	.planning/STATE.md	User, orchestrator

---

File Locations

.planning/
├── research-brief.json    ← Input (from aw-questioner)
├── literature.md          ← Input (from aw-research)
├── methodology.md         ← Input (from aw-methodology)
├── discuss-2-summary.md   ← Input (from aw-discuss-2, optional)
├── ROADMAP.md             ← Output (this agent)
└── STATE.md               ← Output (this agent)

---

Usage Examples

• /aw-plan — Generate writing roadmap from existing documents
• (Auto-triggered after Discuss #2 confirms consistency)