grant-proposal

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Grant Proposal: From Research Ideas to Fundable Application

基金申请书：从研究想法到可获批的申请材料

Draft a grant proposal based on: $ARGUMENTS

基于以下内容起草基金申请书：$ARGUMENTS

Overview

概述

This skill turns validated research ideas into a structured, reviewer-ready grant proposal. It chains sub-skills into a grant-specific pipeline:

/research-lit → /novelty-check → [structure design] → [draft] → /research-review → [revise] → GRANT_PROPOSAL.md
  (survey)      (verify gap)     (aims + matrix)     (prose)    (panel review)     (fix)      (done!)

This is a parallel branch, not part of the linear Workflow 1→1.5→2→3 pipeline. After

/idea-discovery

produces validated ideas, the user can either:

Go to

/experiment-bridge

→

/auto-review-loop

→

/paper-writing

(implement & publish)

Go to
```
/grant-proposal
```
(write funding application first, then implement after funding)

                    ┌→ /experiment-bridge → /auto-review-loop → /paper-writing  (publish track)
/idea-discovery ────┤
                    └→ /grant-proposal → [get funded] → /experiment-bridge → ...  (funding track)

Grant proposals argue for future work (feasibility + potential), not completed work (results + claims). This skill handles the unique requirements of grant writing: narrative arc design, reviewer-facing structure, budget justification, timeline planning, and agency-specific formatting.

本技能可将经过验证的研究想法转化为结构化、符合评审要求的基金申请书。它将多个子技能串联成一条针对基金申请的专属工作流：

/research-lit → /novelty-check → [structure design] → [draft] → /research-review → [revise] → GRANT_PROPOSAL.md
  (文献调研)      (验证研究空白)     (目标+矩阵)     (撰写初稿)    (评审团审核)     (修改)      (完成!)

**这是一条并行分支，不属于线性工作流1→1.5→2→3的一部分。**当

/idea-discovery

生成经过验证的想法后，用户可以选择：

进入

/experiment-bridge

→

/auto-review-loop

→

/paper-writing

（实施研究并发表论文）

进入
```
/grant-proposal
```
（先撰写基金申请，获批后再开展研究）

                    ┌→ /experiment-bridge → /auto-review-loop → /paper-writing  (发表路径)
/idea-discovery ────┤
                    └→ /grant-proposal → [获得资助] → /experiment-bridge → ...  (资助路径)

基金申请书要论证未来工作（可行性+潜力），而非已完成的工作（成果+结论）。本技能针对基金申请的独特需求进行设计：包括叙事结构设计、面向评审的内容框架、预算合理性说明、时间规划以及符合资助机构要求的格式规范。

Constants

常量

GRANT_TYPE =
KAKENHI
— Default grant type. Supported:
```
KAKENHI
```
,
```
NSF
```
,
```
NSFC
```
,
```
ERC
```
,
```
DFG
```
,
```
SNSF
```
,
```
ARC
```
,
```
NWO
```
,
```
GENERIC
```
. Override via argument (e.g.,
```
/grant-proposal "topic — NSF"
```
).
GRANT_SUBTYPE =
auto
— Sub-type within the grant agency. Examples: KAKENHI
```
Start-up
```
/
```
Wakate
```
/
```
Kiban-B
```
; NSFC
```
Youth
```
/
```
Excellent-Youth
```
/
```
Distinguished
```
/
```
Overseas
```
/
```
Key
```
; NSF
```
CAREER
```
/
```
CRII
```
/
```
Standard
```
. Auto-detected from argument or defaults to the most common sub-type.
REVIEWER_MODEL =
gpt-5.4
— Model used via Codex MCP for proposal review. Must be an OpenAI model (e.g.,
```
gpt-5.4
```
,
```
o3
```
,
```
gpt-4o
```
).
OUTPUT_FORMAT =
markdown
— Output format. Supported:
```
markdown
```
,
```
latex
```
. LaTeX uses grant-specific templates when available.
MAX_REVIEW_ROUNDS = 2 — Maximum external review-revise cycles before finalizing.
OUTPUT_DIR =
grant-proposal/
— Directory for generated proposal files.
LANGUAGE =
auto
— Output language. Auto-detected from grant type: KAKENHI→Japanese, NSF→English, NSFC→Chinese, ERC→English, DFG→English (or German), SNSF→English, ARC→English, NWO→English. Override explicitly if needed.
AUTO_PROCEED = false — At each checkpoint, always wait for explicit user confirmation before proceeding. Grant proposals require PI-specific judgment at every stage. Set
```
true
```
only if user explicitly requests fully autonomous mode.

💡 These are defaults. Override by telling the skill, e.g.,
/grant-proposal "topic — NSF CAREER, latex output"
or
/grant-proposal "topic — NSFC Youth, language: English"
.

GRANT_TYPE =
KAKENHI
— 默认基金类型。支持的类型包括：
```
KAKENHI
```
、
```
NSF
```
、
```
NSFC
```
、
```
ERC
```
、
```
DFG
```
、
```
SNSF
```
、
```
ARC
```
、
```
NWO
```
、
```
GENERIC
```
。可通过参数覆盖默认值（例如：
```
/grant-proposal "研究主题 — NSF"
```
）。
GRANT_SUBTYPE =
auto
— 资助机构下属的具体类别。例如：KAKENHI的
```
Start-up
```
/
```
Wakate
```
/
```
Kiban-B
```
；NSFC的
```
Youth
```
/
```
Excellent-Youth
```
/
```
Distinguished
```
/
```
Overseas
```
/
```
Key
```
；NSF的
```
CAREER
```
/
```
CRII
```
/
```
Standard
```
。系统会自动从参数中识别，默认使用最常见的类别。
REVIEWER_MODEL =
gpt-5.4
— 通过Codex MCP用于申请书评审的模型。必须是OpenAI的模型（例如：
```
gpt-5.4
```
、
```
o3
```
、
```
gpt-4o
```
）。
OUTPUT_FORMAT =
markdown
— 输出格式。支持：
```
markdown
```
、
```
latex
```
。当选择LaTeX时，会使用对应基金的专用模板。
MAX_REVIEW_ROUNDS = 2 — 最终定稿前最多可进行的外部评审-修改循环次数。
OUTPUT_DIR =
grant-proposal/
— 生成的申请书文件存储目录。
LANGUAGE =
auto
— 输出语言。系统会根据基金类型自动检测：KAKENHI→日语，NSF→英语，NSFC→中文，ERC→英语，DFG→英语（或德语），SNSF→英语，ARC→英语，NWO→英语。可根据需要手动覆盖默认设置。
AUTO_PROCEED = false — 在每个检查点，必须等待用户明确确认后再继续。基金申请的每个阶段都需要项目负责人（PI）的自主判断。只有当用户明确要求完全自主模式时，才设置为
```
true
```
。

💡 以上为默认设置。可通过告知技能进行覆盖，例如：
/grant-proposal "研究主题 — NSF CAREER, latex输出"
或
/grant-proposal "研究主题 — NSFC Youth, language: English"
。

Grant Type Specifications

基金类型规范

KAKENHI (Japan — JSPS)

KAKENHI（日本 — JSPS）

Field	Detail
Sections	研究目的 (Research Objective), 研究計画・方法 (Plan & Methods), 準備状況 (Preparation Status), 人権の保護 (Ethics, if applicable)
Sub-types	基盤研究 A/B/C (Kiban), 若手研究 (Wakate), 研究活動スタート支援 (Start-up), 国際共同研究 (International), 学術変革領域 (Transformative), 挑戦的研究 (Challenging), DC1/DC2 (doctoral)
Language	Japanese (English technical terms acceptable)
Review criteria	学術的重要性 (academic significance), 独創性 (originality), 研究計画の妥当性 (plan feasibility), 研究遂行能力 (PI capability)
Cultural norms	Explicit yearly milestones (Year 1 / Year 2), budget justification integrated into plan, emphasize 社会的意義 (societal significance), concrete expected outputs (papers, datasets), reference KAKEN database for related funded projects

字段	详情
必填章节	研究目的（Research Objective）、研究計画・方法（研究计划与方法）、準備状況（准备情况）、人権の保護（伦理考量，如适用）
具体类别	基盤研究A/B/C（Kiban）、若手研究（Wakate）、研究活動スタート支援（Start-up）、国際共同研究（国际合作研究）、学術変革領域（学术变革领域）、挑戦的研究（挑战性研究）、DC1/DC2（博士项目）
语言要求	日语（可使用英文专业术语）
评审标准	学術的重要性（学术重要性）、独創性（原创性）、研究計画の妥当性（研究计划可行性）、研究遂行能力（项目负责人执行能力）
文化规范	需明确列出年度里程碑（第1年/第2年），预算说明需融入研究计划，强调社会的意義（社会意义），需列出具体预期产出（论文、数据集），需参考KAKEN数据库中相关已资助项目

NSF (US)

NSF（美国）

Field	Detail
Sections	Project Summary (1p), Project Description (15p max), References Cited, Biographical Sketch, Budget Justification, Data Management Plan
Sub-types	Standard Grant, CAREER (early career), CRII (research initiation), RAPID, EAGER
Language	English
Review criteria	Intellectual Merit, Broader Impacts
Cultural norms	Aim-based structure (Aim 1/2/3), preliminary data strongly expected, broader impacts must be concrete and specific (not generic "benefit society"), Results from Prior Support section

字段	详情
必填章节	项目摘要（1页）、项目说明（最多15页）、参考文献、个人简历、预算说明、数据管理计划
具体类别	标准基金、CAREER（早期职业生涯）、CRII（研究启动）、RAPID、EAGER
语言要求	英语
评审标准	学术价值、广泛影响
文化规范	采用目标导向结构（目标1/2/3），每个目标需提供初步数据，广泛影响部分需具体明确（不能泛泛而谈“造福社会”），需包含“过往资助成果”章节

NSFC (China — 国家自然科学基金)

NSFC（中国 — 国家自然科学基金）

Field	Detail
Sections	立项依据 (Rationale & Significance), 研究内容 (Content), 研究目标 (Objectives), 研究方案 (Plan & Methods), 可行性分析 (Feasibility), 创新性 (Innovation Points), 预期成果 (Expected Outcomes), 研究基础 (PI Foundation & Track Record)
Sub-types	面上项目 (General Program) — emphasis on scientific problem and research accumulation; 青年基金 (Young Scientists Fund) — age ≤35, emphasis on independence and growth potential; 优秀青年基金/优青 (Excellent Young Scientists) — age ≤38, emphasis on outstanding achievements; 杰出青年基金/杰青 (Distinguished Young Scientists) — age ≤45, emphasis on international-leading level; 海外优青 (Overseas Excellent Young Scientists) — emphasis on overseas experience and return contribution plan; 重点项目 (Key Program) — emphasis on systematic in-depth research
Language	Chinese
Review criteria	科学意义 (scientific significance), 创新性 (innovation), 可行性 (feasibility), 研究队伍 (team qualification)
Cultural norms	Heavy emphasis on 国际前沿 (international frontier) positioning, detailed feasibility analysis, explicit citation of applicant's prior publications, 研究基础 section is critical for demonstrating PI capability

字段	详情
必填章节	立项依据（研究背景与意义）、研究内容、研究目标、研究方案、可行性分析、创新性、预期成果、研究基础（项目负责人资历与研究成果）
具体类别	面上项目（General Program）—— 侧重科学问题与研究积累；青年基金（Young Scientists Fund）—— 年龄≤35岁，侧重独立性与成长潜力；优秀青年基金/优青（Excellent Young Scientists）—— 年龄≤38岁，侧重杰出成果；杰出青年基金/杰青（Distinguished Young Scientists）—— 年龄≤45岁，侧重国际领先水平；海外优青（Overseas Excellent Young Scientists）—— 侧重海外经历与归国贡献计划；重点项目（Key Program）—— 侧重系统性深入研究
语言要求	中文
评审标准	科学意义、创新性、可行性、研究团队资质
文化规范	需高度强调国际前沿定位，可行性分析需详尽，需明确引用申请人过往发表的论文，研究基础章节是展示项目负责人能力的关键

ERC (EU — European Research Council)

ERC（欧盟 — 欧洲研究理事会）

Field	Detail
Sections	Extended Synopsis (5p), Scientific Proposal Part B2 (15p)
Sub-types	Starting Grant (2-7 years post-PhD), Consolidator Grant (7-12 years), Advanced Grant (established leaders)
Language	English
Review criteria	Ground-breaking nature, Methodology, PI track record
Cultural norms	Emphasis on "high-risk/high-gain", methodology table with WP/deliverables/milestones, Gantt chart expected, strong PI narrative

字段	详情
必填章节	扩展摘要（5页）、科学提案B2部分（15页）
具体类别	Starting Grant（博士毕业后2-7年）、Consolidator Grant（博士毕业后7-12年）、Advanced Grant（资深研究者）
语言要求	英语
评审标准	突破性、研究方法、项目负责人研究成果
文化规范	需突出“高风险/高回报”的叙事，需包含工作包表格（含交付成果与里程碑），需提供甘特图，需打造有说服力的项目负责人叙事

DFG (Germany — Deutsche Forschungsgemeinschaft)

DFG（德国 — 德意志研究联合会）

Field	Detail
Sections	State of the Art, Objectives, Work Programme, Bibliography, CV
Language	English or German
Review criteria	Scientific quality, Originality, Feasibility, PI qualification

字段	详情
必填章节	研究现状、研究目标、工作计划、参考文献、个人简历
语言要求	英语或德语
评审标准	学术质量、原创性、可行性、项目负责人资质

SNSF (Switzerland — Swiss National Science Foundation)

SNSF（瑞士 — 瑞士国家科学基金会）

Field	Detail
Sections	Summary, Research Plan, Timetable, Budget
Language	English
Review criteria	Scientific relevance, Originality, Feasibility, Track record

字段	详情
必填章节	摘要、研究计划、时间表、预算
语言要求	英语
评审标准	学术相关性、原创性、可行性、研究成果记录

ARC (Australia — Australian Research Council)

ARC（澳大利亚 — 澳大利亚研究理事会）

Field	Detail
Sections	Project Description, Feasibility, Benefit, Budget
Language	English
Review criteria	Research quality, Feasibility, Benefit to Australia

字段	详情
必填章节	项目说明、可行性、社会效益、预算
语言要求	英语
评审标准	研究质量、可行性、对澳大利亚的益处

NWO (Netherlands — Dutch Research Council)

NWO（荷兰 — 荷兰研究理事会）

Field	Detail
Sections	Summary, Proposed Research, Knowledge Utilisation
Language	English
Review criteria	Scientific quality, Innovative character, Knowledge utilisation

字段	详情
必填章节	摘要、拟开展研究、知识应用
语言要求	英语
评审标准	学术质量、创新性、知识应用

GENERIC

GENERIC（通用格式）

For any grant not listed above. User provides section names, page limits, and review criteria via argument:

/grant-proposal "topic — GENERIC, sections: Background|Methods|Impact, language: English"

适用于上述未列出的基金类型。用户可通过参数提供章节名称、页数限制与评审标准：

/grant-proposal "研究主题 — GENERIC, sections: Background|Methods|Impact, language: English"

State Persistence (Compact Recovery)

状态持久化（快速恢复）

Grant proposal drafting is a long task that may trigger context compaction. Persist state to

grant-proposal/GRANT_STATE.json

after each phase:

json

{
  "phase": 2,
  "grant_type": "KAKENHI",
  "grant_subtype": "Start-up",
  "language": "Japanese",
  "codex_thread_id": "019cfcf4-...",
  "gap_statement": "...",
  "aims_count": 3,
  "status": "in_progress",
  "timestamp": "2026-03-18T15:00:00"
}

Write this file at the end of every phase. On invocation, check for this file:

If absent or
```
status: "completed"
```
→ fresh start
If
```
status: "in_progress"
```
and within 24h → resume from saved phase (read
```
GRANT_PROPOSAL.md
```
and
```
GRANT_REVIEW.md
```
to restore context)
If older than 24h → fresh start (stale state)

On completion, set

"status": "completed"

基金申请书起草是一项长期任务，可能会触发上下文压缩。系统会在每个阶段结束后将状态保存至

grant-proposal/GRANT_STATE.json

：

json

{
  "phase": 2,
  "grant_type": "KAKENHI",
  "grant_subtype": "Start-up",
  "language": "Japanese",
  "codex_thread_id": "019cfcf4-...",
  "gap_statement": "...",
  "aims_count": 3,
  "status": "in_progress",
  "timestamp": "2026-03-18T15:00:00"
}

**每个阶段结束后务必写入此文件。**调用技能时，系统会检查该文件：

若文件不存在或
```
status: "completed"
```
→ 全新开始
若
```
status: "in_progress"
```
且时间在24小时内 → 从保存的阶段恢复（读取
```
GRANT_PROPOSAL.md
```
与
```
GRANT_REVIEW.md
```
以恢复上下文）
若超过24小时 → 全新开始（状态已过期）

完成起草后，将

"status": "completed"

。

Workflow

工作流

Phase 0: Input Parsing & Context Gathering

阶段0：输入解析与上下文收集

Parse

$ARGUMENTS

to extract:

Research direction/idea — may reference existing files or be a freeform description
Grant type — detect from keywords (e.g., "科研費"→KAKENHI, "NSF"→NSF, "国自然"→NSFC, "基金"→NSFC)
Grant sub-type — detect from keywords (e.g., "Start-up", "若手", "青年", "CAREER", "优青", "海外优青")
Overrides — output format, language, review rounds

Then gather context from the project directory:

Read
```
IDEA_REPORT.md
```
if it exists (from
```
/idea-discovery
```
)

Read

refine-logs/FINAL_PROPOSAL.md

if it exists (from

/research-refine

)

Read

refine-logs/EXPERIMENT_PLAN.md

if it exists (from

/experiment-plan

)

Read
```
AUTO_REVIEW.md
```
if it exists (from
```
/auto-review-loop
```
— prior review feedback is gold for grants)
Read
```
NARRATIVE_REPORT.md
```
or
```
STORY.md
```
if they exist
Read any existing literature notes or survey documents
Scan for the user's publication list (e.g.,
```
publications.md
```
,
```
cv.md
```
,
```
bio.md
```
,
```
CV.pdf
```
)
Check for
```
grant-proposal/GRANT_STATE.json
```
(resume from prior interrupted run)

If insufficient context exists:

No research idea at all → suggest running
```
/idea-discovery
```
first
No literature survey → will invoke
```
/research-lit
```
inline in Phase 1
No publication list → leave PI qualification section with
```
[TODO: Add publications]
```
placeholders
Has AUTO_REVIEW.md → extract reviewer feedback and use it to strengthen the feasibility narrative

解析

$ARGUMENTS

以提取以下信息：

研究方向/想法 — 可能引用现有文件或为自由文本描述
基金类型 — 从关键词中检测（例如："科研費"→KAKENHI，"NSF"→NSF，"国自然"→NSFC，"基金"→NSFC）
基金具体类别 — 从关键词中检测（例如："Start-up"、"若手"、"青年"、"CAREER"、"优青"、"海外优青"）
覆盖设置 — 输出格式、语言、评审次数

随后从项目目录中收集上下文：

若存在
```
IDEA_REPORT.md
```
则读取（来自
```
/idea-discovery
```
）

若存在

refine-logs/FINAL_PROPOSAL.md

则读取（来自

/research-refine

）

若存在

refine-logs/EXPERIMENT_PLAN.md

则读取（来自

/experiment-plan

）

若存在
```
AUTO_REVIEW.md
```
则读取（来自
```
/auto-review-loop
```
— 过往评审反馈对基金申请至关重要）
若存在
```
NARRATIVE_REPORT.md
```
或
```
STORY.md
```
则读取
读取所有现有的文献笔记或调研文档
扫描用户的发表列表（例如：
```
publications.md
```
、
```
cv.md
```
、
```
bio.md
```
、
```
CV.pdf
```
）
检查是否存在
```
grant-proposal/GRANT_STATE.json
```
（从之前中断的运行中恢复）

若上下文不足：

完全没有研究想法 → 建议先运行
```
/idea-discovery
```
没有文献调研 → 将在阶段1中自动调用
```
/research-lit
```
没有发表列表 → 在项目负责人资质章节留下
```
[TODO: 添加发表论文]
```
占位符
存在AUTO_REVIEW.md → 提取评审反馈并用于强化可行性叙事

Phase 1: Literature & Landscape Positioning

阶段1：文献调研与领域定位

Invoke

/research-lit

to ground the proposal in real literature, then search for competing funded projects:

/research-lit "$ARGUMENTS"

What this does:

Reuse existing surveys if
```
/research-lit
```
was already run and notes exist
Otherwise invoke
```
/research-lit
```
for multi-source literature search (arXiv, Scholar, Zotero, local PDFs)
Search for funded projects in the same area via WebSearch:
- KAKENHI → KAKEN database (https://kaken.nii.ac.jp/)
- NSF → NSF Award Search (https://www.nsf.gov/awardsearch/)
- NSFC → NSFC funded projects
- Other agencies → general web search
Identify competing groups and their recent publications
Run
```
/novelty-check
```
on the proposed research direction to verify the gap is real:
```
/novelty-check "[proposed gap statement]"
```

Build the gap statement — the single most important sentence in the proposal:

"Despite progress in [X], [specific gap] remains unaddressed because [reason].
This proposal addresses this by [approach], which will [expected impact]."

🚦 Checkpoint: Present the landscape summary and gap statement to the user:

📚 Literature & landscape analysis complete:
- [key findings from literature]
- [competing funded projects found]
- Gap statement: "[the gap statement]"

Does this accurately capture the positioning? Should I adjust before designing the proposal structure?

⛔ STOP HERE and wait for user response. Do NOT auto-proceed unless AUTO_PROCEED=true was explicitly set by the user.

Options for the user:

Reply "go" or "ok" → proceed to Phase 2 with current positioning
Reply with adjustments (e.g., "focus more on X", "the gap should emphasize Y") → refine and re-present
Reply "stop" → end the skill, save current progress to
```
grant-proposal/DRAFT_NOTES.md
```

State: Write

GRANT_STATE.json

with

phase: 1

and the gap statement.

调用

/research-lit

使申请书基于真实文献，然后搜索已资助的同类项目：

/research-lit "$ARGUMENTS"

此步骤的作用：

若
```
/research-lit
```
已运行且存在笔记，则复用现有调研结果
否则调用
```
/research-lit
```
进行多来源文献搜索（arXiv、Scholar、Zotero、本地PDF）
通过网络搜索已资助项目：
- KAKENHI → KAKEN数据库（https://kaken.nii.ac.jp/）
- NSF → NSF资助项目搜索（https://www.nsf.gov/awardsearch/）
- NSFC → NSFC已资助项目
- 其他机构 → 通用网络搜索
识别竞争团队及其近期发表的论文
对拟开展的研究方向运行
```
/novelty-check
```
以验证研究空白的真实性：
```
/novelty-check "[拟提出的研究空白说明]"
```

构建研究空白说明 — 这是申请书中最重要的一句话：

"尽管[X领域]已取得进展，但[具体研究空白]仍未得到解决，原因在于[具体原因]。本申请将通过[研究方法]解决这一问题，预期将产生[预期影响]。"

🚦 检查点： 向用户展示领域调研总结与研究空白说明：

📚 文献与领域分析完成：
- [文献调研关键发现]
- [找到的同类已资助项目]
- 研究空白说明："[研究空白说明内容]"

以上内容是否准确反映了研究定位？是否需要在设计申请书结构前进行调整？

⛔ 在此处停止，等待用户回复。 除非用户明确设置AUTO_PROCEED=true，否则不得自动继续。

用户可选择：

回复**"go"或"ok"** → 基于当前定位进入阶段2
回复调整建议（例如："更多关注X领域"、"研究空白应强调Y点"） → 优化后重新展示
回复**"stop"** → 结束技能，将当前进度保存至
```
grant-proposal/DRAFT_NOTES.md
```

状态保存：将

GRANT_STATE.json

写入，设置

phase: 1

并保存研究空白说明。

Phase 2: Narrative Structure & Aims Design

阶段2：叙事结构与目标设计

Design the proposal's logical architecture before writing any prose.

在撰写初稿前，先设计申请书的逻辑架构。

2.1 Define Specific Aims (2-4)

2.1 确定具体研究目标（2-4个）

Each aim must satisfy:

Independently valuable — if one aim fails, others still produce publishable results
Logically connected — Aim 1 enables Aim 2, Aim 2 informs Aim 3
Concrete deliverables — each aim maps to specific outputs (papers, datasets, tools, benchmarks)
Feasible within budget and timeline

每个目标需满足：

独立价值 — 若某个目标失败，其他目标仍能产生可发表的成果
逻辑关联 — 目标1支撑目标2，目标2为目标3提供信息
具体交付成果 — 每个目标对应具体产出（论文、数据集、工具、基准测试）
在预算与时间范围内可行

2.2 Build Claims-Aims-Evidence Matrix

2.2 构建主张-目标-证据矩阵

markdown

| Aim | Key Claim | Preliminary Evidence | Proposed Validation | Risk Level | Deliverable |
|-----|-----------|---------------------|--------------------|-----------:|-------------|
| Aim 1 | [claim] | [pilot data, prior work] | [experiments] | LOW | [paper, dataset] |
| Aim 2 | [claim] | [theoretical basis] | [experiments] | MEDIUM | [paper, tool] |

markdown

| 目标 | 核心主张 | 初步证据 | 拟开展的验证工作 | 风险等级 | 交付成果 |
|-----|-----------|---------------------|--------------------|-----------:|-------------|
| 目标1 | [主张内容] | [试点数据、过往工作] | [实验方案] | 低 | [论文、数据集] |
| 目标2 | [主张内容] | [理论基础] | [实验方案] | 中 | [论文、工具] |

2.3 Design the Narrative Arc

2.3 设计叙事结构

Grant proposals follow a fundamentally different arc from papers:

Problem → Why Now → What We Propose → Why It Will Work → What We Will Deliver
         (not: Problem → Method → Results → Implications)

Problem: What gap exists and why it matters (scientific + societal)
Why Now: What recent developments make this the right time (new data, new methods, new need)
What We Propose: The specific aims and approach
Why It Will Work: Preliminary data, PI track record, team expertise, feasibility arguments
What We Will Deliver: Concrete outputs, timeline, expected publications

基金申请书的叙事结构与论文有本质区别：

问题 → 为何现在开展 → 我们的方案 → 为何可行 → 我们将交付什么
         （而非：问题 → 方法 → 结果 → 意义）

问题：存在什么研究空白，以及其重要性（科学+社会层面）
为何现在开展：近期哪些进展使得此时开展研究恰逢其时（新数据、新方法、新需求）
我们的方案：具体研究目标与方法
为何可行：初步数据、项目负责人研究成果、团队专业能力、可行性论证
我们将交付什么：具体产出、时间表、预期发表论文

2.4 Timeline & Milestones

2.4 时间表与里程碑

Design year-by-year (or quarter-by-quarter) plan:

markdown

undefined

设计按年（或按季度）划分的计划：

markdown

undefined

Year 1

第1年

Q1-Q2: [Aim 1 tasks]
Q3-Q4: [Aim 1 completion + Aim 2 start]
Expected outputs: [papers, datasets]

Q1-Q2: [目标1任务]
Q3-Q4: [完成目标1 + 启动目标2]
预期产出: [论文、数据集]

Year 2

第2年

Q1-Q2: [Aim 2 completion + Aim 3]
Q3-Q4: [Aim 3 completion + synthesis]
Expected outputs: [papers, tools, final report]

undefined

Q1-Q2: [完成目标2 + 启动目标3]
Q3-Q4: [完成目标3 + 综合分析]
预期产出: [论文、工具、最终报告]

undefined

2.5 Structural Review

2.5 结构评审

Invoke

/research-review

to get critical feedback on the proposal structure before drafting:

/research-review "[GRANT_TYPE] [GRANT_SUBTYPE] proposal structure:
Gap: [gap statement]
Aims: [aims list with claims-evidence matrix]
Timeline: [timeline]
— reviewer persona: [GRANT_TYPE] review panelist"

What this does:

GPT-5.4 xhigh acts as a grant review panelist (not a paper reviewer)
Evaluates aims independence, narrative arc, risk identification, timeline realism
Identifies the single biggest reviewer concern
Provides actionable fixes ranked by severity

Apply structural feedback before proceeding to drafting.

🚦 Checkpoint: Present the proposal structure to the user:

🏗️ Proposal structure designed:
- Gap: [gap statement]
- Aim 1: [title] — Risk: LOW
- Aim 2: [title] — Risk: MEDIUM
- Aim 3: [title] — Risk: LOW
- Timeline: [summary]
- Reviewer feedback: [key points from GPT-5.4]

Proceed to section drafting? Or adjust the structure?

⛔ STOP HERE. This is the most critical checkpoint — the proposal structure determines everything downstream.

Options for the user:

Reply "go" or "ok" → proceed to Phase 3 (section drafting)
Reply with structural changes (e.g., "merge Aim 2 and 3", "add an aim about X", "reduce to 2 aims") → redesign and re-present
Reply "back" → return to Phase 1 to adjust the gap/positioning
Reply "stop" → save current structure to
```
grant-proposal/DRAFT_NOTES.md
```

State: Write

GRANT_STATE.json

with

phase: 2

, aims summary, and Codex threadId.

在撰写初稿前，调用

/research-review

获取对申请书结构的关键反馈：

/research-review "[GRANT_TYPE] [GRANT_SUBTYPE] 申请书结构：
研究空白：[研究空白说明]
目标：[带主张-证据矩阵的目标列表]
时间表：[时间表摘要]
— 评审者角色：[GRANT_TYPE]评审团成员"

此步骤的作用：

GPT-5.4 xhigh扮演基金评审团成员（而非论文评审）
评估目标独立性、叙事结构、风险识别、时间表合理性
识别评审者最关注的核心问题
按优先级提供可操作的修改建议

在进入撰写阶段前应用结构反馈。

🚦 检查点： 向用户展示申请书结构：

🏗️ 申请书结构已设计完成：
- 研究空白：[研究空白说明]
- 目标1：[标题] — 风险等级：低
- 目标2：[标题] — 风险等级：中
- 目标3：[标题] — 风险等级：低
- 时间表：[摘要]
- 评审反馈：[GPT-5.4的关键反馈要点]

是否进入章节撰写阶段？还是需要调整结构？

⛔ 在此处停止。这是最关键的检查点 — 申请书结构决定了后续所有内容。

用户可选择：

回复**"go"或"ok"** → 进入阶段3（章节撰写）
回复结构修改建议（例如："合并目标2和3"、"添加关于X的目标"、"减少至2个目标"） → 重新设计并展示
回复**"back"** → 返回阶段1调整研究空白/定位
回复**"stop"** → 将当前结构保存至
```
grant-proposal/DRAFT_NOTES.md
```

状态保存：将

GRANT_STATE.json

写入，设置

phase: 2

，保存目标摘要与Codex线程ID。

Phase 3: Section Drafting

阶段3：章节撰写

Draft each section according to the grant type template. Write complete prose, not outlines or placeholders.

What this does:

Writes all required sections in the agency-specific language and tone
Pulls content from IDEA_REPORT.md, FINAL_PROPOSAL.md, and literature notes
Uses
```
/paper-illustration
```
for figure generation (if user requests)
Leaves
```
[TODO]
```
only for PI-specific information,
```
[AMOUNT]
```
for budget figures
Outputs
```
grant-proposal/GRANT_PROPOSAL.md
```

根据对应基金类型的模板撰写每个章节。撰写完整的正文内容，而非大纲或占位符。

此步骤的作用：

以符合资助机构要求的语言与语气撰写所有必填章节
从IDEA_REPORT.md、FINAL_PROPOSAL.md与文献笔记中提取内容
若用户要求，调用
```
/paper-illustration
```
生成图表
仅针对项目负责人专属信息留下
```
[TODO]
```
占位符，预算金额留为
```
[AMOUNT]
```
占位符
输出至
```
grant-proposal/GRANT_PROPOSAL.md
```

Drafting Order (optimized for narrative coherence)

撰写顺序（优化叙事连贯性）

Specific Aims / Research Objective — the "abstract" of the grant. Write first, refine last.
Background / Significance / State of the Art — establish the problem and gap.
Research Plan / Methods — per aim, with feasibility arguments.
Figures — generate key diagrams (see below).
Timeline & Milestones — year-by-year deliverables.
PI Qualification / Preparation Status — track record, team, infrastructure.
Budget Justification — narrative only (leave dollar/yen amounts as
```
[AMOUNT]
```
placeholders).
Broader Impacts / Societal Significance — if required by the grant type.

具体目标 / 研究目的 — 基金申请书的“摘要”。先撰写，最后再优化。
背景 / 研究意义 / 研究现状 — 阐述问题与研究空白。
研究计划 / 研究方法 — 按目标展开，包含可行性论证。
图表 — 生成关键示意图（如下文所述）。
时间表与里程碑 — 按年划分的交付成果。
项目负责人资质 / 准备情况 — 研究成果记录、团队、基础设施。
预算说明 — 仅撰写叙事性说明（具体金额留为
```
[AMOUNT]
```
占位符）。
广泛影响 / 社会意义 — 若基金类型要求。

Figure Generation

图表生成

Grant proposals benefit greatly from clear diagrams. Generate the following figures using SVG or matplotlib (save to

grant-proposal/figures/

全体構成図 / Overview Diagram — Show the relationship between aims (Aim 1 → Aim 2 → Aim 3), shared resources (participants, stimuli, pipeline), and outputs. This is the single most important figure.
実験パラダイム図 / Experimental Paradigm — Visual schematic of each paradigm (stimulus timing, conditions, EEG recording).
年次計画 / Timeline Gantt Chart — Year-by-year (or H1/H2) milestones with deliverables.

For AI-generated publication-quality figures, invoke

/paper-illustration

/paper-illustration "Overview diagram showing [aims relationship + shared resources] for grant proposal"

For simpler diagrams (flowcharts, Gantt charts), generate clean SVG or matplotlib directly via code.

🚦 Figure Checkpoint: Before generating, ask which figures the user wants:

🎨 The following figures would strengthen this proposal:
1. 全体構成図 / Overview — aims relationship + shared resources
2. 実験パラダイム図 / Paradigm — stimulus timing + conditions
3. 年次計画 / Gantt — timeline with milestones

Which should I generate? (e.g., "1 and 3", "all", "skip")

⛔ Wait for user response. Generate only the requested figures.

清晰的图表能显著提升申请书质量。使用SVG或matplotlib生成以下图表（保存至

grant-proposal/figures/

）：

全体構成図 / 总览图 — 展示目标间的关系（目标1→目标2→目标3）、共享资源（参与者、刺激、工作流）与产出。这是最重要的图表。
実験パラダイム図 / 实验范式图 — 每个实验范式的可视化示意图（刺激时序、条件、EEG记录）。
年次計画 / 甘特图 — 按年（或上半年/下半年）划分的里程碑与交付成果。

若需AI生成符合出版质量的图表，调用

/paper-illustration

：

/paper-illustration "基金申请总览图，展示[目标关系 + 共享资源]"

对于简单图表（流程图、甘特图），可直接通过代码生成简洁的SVG或matplotlib图表。

🚦 图表检查点： 生成前询问用户需要哪些图表：

🎨 以下图表可提升本申请书的质量：
1. 全体構成図 / 总览图 — 目标关系 + 共享资源
2. 実験パラダイム図 / 实验范式图 — 刺激时序 + 实验条件
3. 年次計画 / 甘特图 — 含里程碑的时间表

需要生成哪些图表？（例如："1和3"、"全部"、"跳过"）

⛔ 等待用户回复。 仅生成用户要求的图表。

Grant-Specific Drafting Guidelines

基金专属撰写指南

KAKENHI:

Write in formal Japanese academic style (である調, not です/ます調)
Use 「」for Japanese quotations, bold for emphasis
Structure: 研究の学術的背景 → 研究期間内に何をどこまで明らかにするか → 本研究の学術的な特色・独創性
Include explicit 年次計画 (yearly plan) with concrete milestones
Emphasize 社会的意義 (societal significance)
Reference related KAKEN-funded projects to show awareness of the field

NSF:

Write in clear, direct English
Use Aim-based structure with bold headings
Preliminary data paragraphs for each Aim (with figure references)
Broader Impacts must be concrete: specific outreach activities, broadening participation plans
Include Results from Prior Support (if PI has prior NSF funding)

NSFC:

Write in formal Chinese academic style
立项依据 must position work at 国际前沿 (international frontier)
创新性 section must list numbered innovation points (创新点)
研究基础 must cite PI's own publications (with IF and citations if possible)
可行性分析 must address: technical feasibility, team capability, time feasibility, equipment/conditions

ERC:

Write a compelling "high-risk/high-gain" narrative
Extended Synopsis must be self-contained and compelling
Include Work Package table with deliverables and milestones
Gantt chart (describe in text, or generate as figure)

KAKENHI：

使用正式的日语学术文体（である调，而非です/ます调）
日语引用使用「」，强调内容使用粗体
结构：研究的学术背景 → 研究期间内将明确什么问题 → 本研究的学术特色与原创性
需包含明确的年次計画（年度计划）与具体里程碑
强调社会的意義（社会意义）
引用相关KAKEN资助项目以展示对领域的了解

NSF：

使用清晰、直接的英语
采用目标导向结构，使用粗体标题
每个目标需包含初步数据段落（含图表引用）
广泛影响部分需具体：明确的推广活动、扩大参与计划
若项目负责人曾获得NSF资助，需包含“过往资助成果”章节

NSFC：

使用正式的中文学术文体
立项依据需将研究定位在国际前沿
创新性章节需列出编号的创新点
研究基础需引用项目负责人本人的发表论文（若可能，需包含影响因子与引用量）
可行性分析需涵盖：技术可行性、团队能力、时间可行性、设备/条件

ERC：

撰写引人入胜的“高风险/高回报”叙事
扩展摘要需独立完整且有说服力
需包含工作包表格，含交付成果与里程碑
甘特图（可在文本中描述，或生成为图表）

For Each Section

每个章节的撰写要求

Pull relevant content from IDEA_REPORT.md, FINAL_PROPOSAL.md, literature notes
Write complete prose — no
```
[TODO]
```
except for PI-specific information
Include figure/table placeholders where appropriate (e.g.,
```
[Figure 1: System architecture]
```
)
Cite references properly — use citation keys, will build bibliography later
Match the agency's tone and style — formal Japanese for KAKENHI, direct English for NSF, etc.

提取相关内容 从IDEA_REPORT.md、FINAL_PROPOSAL.md与文献笔记中提取
撰写完整正文 — 除项目负责人专属信息外，不得留下
```
[TODO]
```
添加图表/表格占位符 （例如：
```
[图1：系统架构]
```
）
正确引用参考文献 — 使用引用标识，后续将生成参考文献列表
匹配资助机构的语气与风格 — KAKENHI使用正式日语，NSF使用直接英语，以此类推

Phase 4: External Review

阶段4：外部评审

Invoke

/research-review

on the complete draft for grant-type-specific evaluation:

/research-review "Complete [GRANT_TYPE] [GRANT_SUBTYPE] proposal draft. Evaluate as a [GRANT_TYPE] review panelist using official criteria. [PASTE FULL PROPOSAL TEXT]"

What this does:

GPT-5.4 xhigh acts as a grant review panelist
Scores each section 1-5 using agency-specific criteria
Identifies fatal flaws and recommends funding/revisions/rejection
Provides ranked action items for improvement
All feedback saved to
```
grant-proposal/GRANT_REVIEW.md
```

⚠️ Codex MCP fallback: If
mcp__codex__codex
is not available (no OpenAI API key), skip external review. Note "External review skipped — no Codex MCP available. Consider running
/auto-review-loop-llm
separately." in GRANT_REVIEW.md. The proposal is still usable without external review.

/research-review

is invoked (preferred), it handles the Codex call internally. If calling Codex directly (e.g., to maintain thread context from Phase 2):

调用

/research-review

对完整初稿进行针对基金类型的评估：

/research-review "完整的[GRANT_TYPE] [GRANT_SUBTYPE]申请书初稿。请以[GRANT_TYPE]评审团成员的身份，使用官方标准进行评估。[粘贴完整申请书文本]"

此步骤的作用：

GPT-5.4 xhigh扮演基金评审团成员
使用对应机构的标准为每个章节打分（1-5分）
识别致命缺陷并建议资助/修改/拒绝
按优先级提供可操作的改进建议
所有反馈保存至
```
grant-proposal/GRANT_REVIEW.md
```

⚠️ Codex MCP备选方案：若
mcp__codex__codex
不可用（无OpenAI API密钥），则跳过外部评审。在GRANT_REVIEW.md中注明“跳过外部评审 — 无Codex MCP可用。建议单独运行
/auto-review-loop-llm
。”。即使没有外部评审，申请书仍可使用。

若调用

/research-review

（推荐方式），它会在内部处理Codex调用。若直接调用Codex（例如：为了保持阶段2的线程上下文）：

Round 1 (full draft review):

第1轮（完整初稿评审）：

mcp__codex__codex-reply:
  threadId: [from Phase 2]
  config: {"model_reasoning_effort": "xhigh"}
  prompt: |
    Review this complete [GRANT_TYPE] [GRANT_SUBTYPE] proposal draft.

    Act as a [GRANT_TYPE] review panelist. Evaluate using the official criteria:

    [INSERT GRANT-TYPE-SPECIFIC CRITERIA — see Grant Type Specifications above]

    For each section:
    1. Score 1-5 (5 = excellent)
    2. Strongest aspect
    3. Most critical weakness
    4. Specific fix suggestion (actionable, not vague)

    Overall assessment:
    - Would you recommend funding? (Yes / Yes with revisions / No)
    - Single most impactful change to improve funding chances?
    - Any fatal flaws?

    [PASTE FULL PROPOSAL TEXT]

mcp__codex__codex-reply:
  threadId: [来自阶段2]
  config: {"model_reasoning_effort": "xhigh"}
  prompt: |
    评审此完整的[GRANT_TYPE] [GRANT_SUBTYPE]申请书初稿。

    请以[GRANT_TYPE]评审团成员的身份，使用以下官方标准进行评估：

    [插入基金类型专属标准 — 见上文基金类型规范]

    针对每个章节：
    1. 打分1-5（5=优秀）
    2. 最突出的优点
    3. 最关键的不足
    4. 具体的修改建议（可操作，而非模糊表述）

    总体评估：
    - 是否建议资助？（是 / 是，但需修改 / 否）
    - 提升获批几率最有效的一项修改是什么？
    - 是否存在致命缺陷？

    [粘贴完整申请书文本]

Round 2+ (after revisions):

第2轮及以后（修改后评审）：

If MAX_REVIEW_ROUNDS > 1 and revisions were applied:

mcp__codex__codex-reply:
  threadId: [saved from Round 1]
  config: {"model_reasoning_effort": "xhigh"}
  prompt: |
    [Round N review of revised [GRANT_TYPE] [GRANT_SUBTYPE] proposal]

    Since your last review, I have applied the following changes:
    1. [Change 1]: [what was done]
    2. [Change 2]: [what was done]
    3. [Change 3]: [what was done]

    Please re-evaluate. Same format: section scores, overall assessment, remaining weaknesses.
    Focus on whether the CRITICAL and MAJOR issues from Round 1 have been adequately addressed.

    [PASTE REVISED PROPOSAL TEXT]

若MAX_REVIEW_ROUNDS > 1且已应用修改：

mcp__codex__codex-reply:
  threadId: [第1轮保存的ID]
  config: {"model_reasoning_effort": "xhigh"}
  prompt: |
    [第N轮评审：修改后的[GRANT_TYPE] [GRANT_SUBTYPE]申请书]

    自上次评审后，已做出以下修改：
    1. [修改1]：[具体内容]
    2. [修改2]：[具体内容]
    3. [修改3]：[具体内容]

    请重新评估。使用相同格式：章节打分、总体评估、剩余不足。
    重点关注第1轮中指出的关键与主要问题是否已得到充分解决。

    [粘贴修改后的申请书文本]

Phase 5: Revision & Output

阶段5：修改与输出

5.1 Apply Reviewer Feedback

5.1 应用评审反馈

Parse reviewer feedback into severity levels:

CRITICAL — fatal flaws that would lead to rejection. Fix immediately.
MAJOR — significant weaknesses. Fix before submission.
MINOR — suggestions for improvement. Fix if time allows.

Implement CRITICAL and MAJOR fixes. If MAX_REVIEW_ROUNDS > 1, re-submit for another round via

mcp__codex__codex-reply

将评审反馈按严重程度分类：

CRITICAL（关键） — 会导致申请被拒绝的致命缺陷。需立即修复。
MAJOR（主要） — 显著不足。需在提交前修复。
MINOR（次要） — 改进建议。若时间允许则修复。

修复关键与主要缺陷。若MAX_REVIEW_ROUNDS > 1，通过

mcp__codex__codex-reply

提交再次评审。

5.2 Generate Output

5.2 生成输出

Markdown output (default):

grant-proposal/
├── GRANT_PROPOSAL.md          # Complete proposal, all sections
├── GRANT_REVIEW.md            # Review history and reviewer feedback
├── GRANT_STATE.json           # State persistence file
├── figures/                   # Generated diagrams (if any)
└── references.bib             # Bibliography (if citations were used)

LaTeX output (when OUTPUT_FORMAT = latex):

grant-proposal/
├── main.tex                   # Master file
├── sections/
│   ├── aims.tex               # Specific Aims / Research Objective
│   ├── background.tex         # Background / Significance
│   ├── research_plan.tex      # Research Plan / Methods
│   ├── timeline.tex           # Timeline & Milestones
│   ├── pi_qualification.tex   # PI Qualification / Track Record
│   └── budget.tex             # Budget Justification (if applicable)
├── references.bib
└── figures/                   # Any generated diagrams

Markdown输出（默认）：

grant-proposal/
├── GRANT_PROPOSAL.md          # 完整申请书，包含所有章节
├── GRANT_REVIEW.md            # 评审历史与评审反馈
├── GRANT_STATE.json           # 状态持久化文件
├── figures/                   # 生成的示意图（若有）
└── references.bib             # 参考文献列表（若使用引用）

LaTeX输出（当OUTPUT_FORMAT = latex时）：

grant-proposal/
├── main.tex                   # 主文件
├── sections/
│   ├── aims.tex               # 具体目标 / 研究目的
│   ├── background.tex         # 背景 / 研究意义
│   ├── research_plan.tex      # 研究计划 / 研究方法
│   ├── timeline.tex           # 时间表与里程碑
│   ├── pi_qualification.tex   # 项目负责人资质 / 研究成果记录
│   └── budget.tex             # 预算说明（若适用）
├── references.bib
└── figures/                   # 生成的示意图（若有）

5.3 Final Checks

5.3 最终检查

Before declaring done:

All sections required by the grant type are present and complete
Gap statement is clear and appears early in the proposal
Each aim is independently valuable and logically connected
Timeline includes concrete yearly milestones and deliverables
PI qualification section has content (or clear
```
[TODO]
```
placeholders)
Budget justification uses
```
[AMOUNT]
```
placeholders (no fabricated numbers)
Language matches the grant type (Japanese for KAKENHI, Chinese for NSFC, etc.)
No leftover
```
[TODO]
```
markers except for PI-specific information
References are real (no hallucinated citations)
Review feedback has been addressed (CRITICAL and MAJOR items)

🚦 Final Checkpoint: Present the completed proposal summary:

📝 Grant proposal draft complete:
- Type: [GRANT_TYPE] [GRANT_SUBTYPE]
- Language: [language]
- Aims: [N] aims covering [summary]
- Timeline: [N] years
- Review score: [summary from GPT-5.4]
- Output: grant-proposal/GRANT_PROPOSAL.md

Files saved to grant-proposal/. Please review and customize:
1. PI qualification section (add your publications and track record)
2. Budget amounts (replace [AMOUNT] placeholders)
3. Any [TODO] markers for personal information

What would you like to do next?
- "figures" → generate proposal diagrams
- "review again" → run another round of external review
- "latex" → convert to LaTeX format
- "done" → finalize

在宣布完成前，需完成以下检查：

基金类型要求的所有章节均已存在且完整
研究空白说明清晰且出现在申请书前部
每个目标具有独立价值且逻辑关联
时间表包含具体的年度里程碑与交付成果
项目负责人资质章节有内容（或明确的
```
[TODO]
```
占位符）
预算说明仅为叙事性内容，具体金额留为
```
[AMOUNT]
```
占位符
语言符合基金类型要求（KAKENHI→日语，NSFC→中文等）
除项目负责人专属信息外，无剩余
```
[TODO]
```
标记
参考文献真实（无虚构引用）
评审反馈已得到处理（关键与主要问题）

🚦 最终检查点： 向用户展示已完成的申请书摘要：

📝 基金申请书初稿已完成：
- 类型：[GRANT_TYPE] [GRANT_SUBTYPE]
- 语言：[语言]
- 目标：[N]个目标，涵盖[摘要]
- 时间表：[N]年
- 评审得分：[GPT-5.4的摘要]
- 输出文件：grant-proposal/GRANT_PROPOSAL.md

文件已保存至grant-proposal/。请审阅并自定义以下内容：
1. 项目负责人资质章节（添加你的发表论文与研究成果记录）
2. 预算金额（替换`[AMOUNT]`占位符）
3. 任何个人信息的`[TODO]`标记

接下来你想做什么？
- "figures" → 生成申请书示意图
- "review again" → 再进行一轮外部评审
- "latex" → 转换为LaTeX格式
- "done" → 最终定稿

Key Rules

核心规则

Large file handling: If the Write tool fails due to file size, immediately retry using Bash (
```
cat << 'EOF' > file
```
) to write in chunks. Do NOT ask the user for permission — just do it silently.
Do NOT fabricate budget amounts. Generate narrative budget justification only. Leave specific dollar/yen/yuan/euro amounts as
```
[AMOUNT]
```
placeholders for the user to fill in.
Do NOT fabricate PI information. If no publication list is available, leave
```
[TODO: Add publications]
```
placeholders. Never invent papers, grants, or credentials.
Do NOT hallucinate citations. Use references from literature survey. Mark uncertain citations with
```
[VERIFY]
```
.
Grant ≠ paper. A grant argues for future work (feasibility + potential). A paper argues for completed work (results + claims). Write accordingly — emphasize "what we will do" and "why it will work", not "what we found."
Aims must be independently valuable. If Aim 2 fails, Aim 1 and Aim 3 should still produce publishable results.
Preliminary data de-risks. Include any pilot results, existing datasets, or prior publications that demonstrate feasibility.
Reviewer-facing structure. Bold key sentences. Use numbered lists for clarity. Make the reviewer's job easy.
Cultural norms matter. KAKENHI expects 社会的意義; NSF expects Broader Impacts; NSFC expects 国际前沿 positioning. Missing these is a red flag for reviewers.
Feishu notifications are optional. If
```
~/.claude/feishu.json
```
exists, send
```
checkpoint
```
at each phase transition and
```
pipeline_done
```
at final output. If absent, skip silently.

大文件处理：若Write工具因文件大小失败，立即使用Bash重试（
```
cat << 'EOF' > file
```
）分块写入。无需询问用户许可 — 直接静默执行。
不得虚构预算金额。仅生成叙事性预算说明。具体金额（美元/日元/人民币/欧元）留为
```
[AMOUNT]
```
占位符，由用户自行填写。
不得虚构项目负责人信息。若没有发表列表，留下
```
[TODO: 添加发表论文]
```
占位符。绝不能虚构论文、基金或资质。
不得虚构引用。使用文献调研中的参考文献。对不确定的引用标记
```
[VERIFY]
```
。
基金≠论文。基金论证未来工作（可行性+潜力），论文论证已完成工作（成果+结论）。需据此调整撰写方式 — 强调“我们将做什么”与“为何可行”，而非“我们发现了什么”。
目标必须具有独立价值。若目标2失败，目标1与目标3仍需能产生可发表的成果。
初步数据降低风险。需包含任何试点结果、现有数据集或过往发表论文以证明可行性。
面向评审者的结构。关键句子使用粗体。使用编号列表提升清晰度。让评审者的工作更轻松。
文化规范至关重要。KAKENHI要求社会的意義；NSF要求广泛影响；NSFC要求国际前沿定位。缺失这些内容会引起评审者的警惕。
飞书通知为可选功能。若存在
```
~/.claude/feishu.json
```
，则在每个阶段转换时发送
```
checkpoint
```
通知，在最终输出时发送
```
pipeline_done
```
通知。若不存在则静默跳过。

Parameter Pass-Through

参数传递

Parameters can be passed inline with

—

separator. They flow to sub-skills when invoked:

/grant-proposal "topic — KAKENHI Start-up, sources: zotero, arxiv download: true"

Parameter	Default	Description	Passed to
`grant type`	KAKENHI	Agency (KAKENHI/NSF/NSFC/ERC/DFG/SNSF/ARC/NWO/GENERIC)	—
`grant subtype`	auto	Sub-type (Start-up/Wakate/CAREER/Youth/etc.)	—
`output format`	markdown	`markdown` or `latex`	—
`language`	auto	Output language override	—
`max review rounds`	2	External review cycles	—
`sources`	all	Literature sources	→ `/research-lit`
`arxiv download`	false	Download arXiv PDFs	→ `/research-lit`
`reviewer model`	gpt-5.4	Codex review model	→ Codex MCP
`auto proceed`	false	Skip checkpoints	—

可使用

—

分隔符在调用时传递参数。这些参数会在调用子技能时传递：

/grant-proposal "研究主题 — KAKENHI Start-up, sources: zotero, arxiv download: true"

参数	默认值	说明	传递至
`grant type`	KAKENHI	资助机构（KAKENHI/NSF/NSFC/ERC/DFG/SNSF/ARC/NWO/GENERIC）	—
`grant subtype`	auto	具体类别（Start-up/Wakate/CAREER/Youth等）	—
`output format`	markdown	`markdown` 或 `latex`	—
`language`	auto	输出语言覆盖设置	—
`max review rounds`	2	外部评审循环次数	—
`sources`	all	文献来源	→ `/research-lit`
`arxiv download`	false	下载arXiv PDF	→ `/research-lit`
`reviewer model`	gpt-5.4	Codex评审模型	→ Codex MCP
`auto proceed`	false	跳过检查点	—

Composing with Other Skills

与其他技能的组合

Sub-skills used by this skill

本技能使用的子技能

Sub-skill	Phase	Purpose
`/research-lit`	1	Literature survey (if not already done)
`/novelty-check`	1	Verify the gap is real
`/research-review`	2, 4	Structural review + full draft review
`/paper-illustration`	3	Generate proposal figures (optional)

子技能	阶段	用途
`/research-lit`	1	文献调研（若未完成）
`/novelty-check`	1	验证研究空白的真实性
`/research-review`	2, 4	结构评审 + 完整初稿评审
`/paper-illustration`	3	生成申请书示意图（可选）

Funding Track (this skill's primary use case)

资助路径（本技能的主要使用场景）

/idea-discovery "direction"              ← Workflow 1: find validated ideas
/research-refine "idea"                  ← sharpen the method
/grant-proposal "idea — KAKENHI"         ← this skill: write the grant proposal
                                          ← [submit & get funded]
/experiment-bridge                       ← implement experiments with funding
/auto-review-loop "results"              ← Workflow 2: iterate until submission-ready
/paper-writing                           ← Workflow 3: write the paper

/idea-discovery "研究方向"              ← 工作流1：找到经过验证的想法
/research-refine "研究想法"                  ← 优化研究方法
/grant-proposal "研究想法 — KAKENHI"         ← 本技能：撰写基金申请书
                                          ← [提交申请并获得资助]
/experiment-bridge                       ← 利用资助开展实验
/auto-review-loop "实验结果"              ← 工作流2：迭代至可提交状态
/paper-writing                           ← 工作流3：撰写论文

Publish Track (skip this skill)

发表路径（跳过本技能）

/idea-discovery → /experiment-bridge → /auto-review-loop → /paper-writing → submit

/idea-discovery → /experiment-bridge → /auto-review-loop → /paper-writing → 提交