spread1000-proposal-writer

name	spread1000-proposal-writer
description	Draft the SPReAD proposal (research plan document). Integrate the research plan, Azure architecture design, and cost estimate to generate a proposal aligned with the 公募要領 requirements. Use when writing the proposal, finalizing the research plan document, or preparing submission documents.

Proposal Writer

Integrate the research plan, Azure architecture design, and cost estimate to generate a proposal aligned with SPReAD 公募要領 requirements.

Use This Skill When

• Research plan, Azure architecture, and cost estimate are ready, and it is time to draft the proposal
• Finalizing the research plan document (研究計画書) per SPReAD 公募要領
• Performing quality checks or refining the proposal

Required Inputs

• output/{project-name}/phase0-research-plan.md (research plan)
• output/{project-name}/phase1-azure-architecture.md (Azure architecture design)
• output/{project-name}/phase2-cost-estimate.md (cost estimate)
• Principal investigator / co-investigator information
• Affiliated institution information
• Research period: from the grant decision date to January 6, Reiwa 9 (~180 days)

Workflow

1. Prerequisite check (MANDATORY):
   • Confirm that output/{project-name}/phase2-cost-estimate.md exists
   • If it does not exist, run spread1000-cost-estimator first to generate the cost estimate before proceeding with proposal drafting
   • If Phase 2 is in "Draft (prices unverified)" state, display a bold warning ⚠️ 価格未検証（推定値） in the expense section
   • Do NOT fill expense figures using LLM memory/estimated values. Always use the retrieved unit prices from Phase 2
2. Input integration: Read and integrate deliverables from Phases 0–2
3. Review 公募要領 requirements: Confirm SPReAD 公募要領 requirements
   • Read references/proposal-guidelines.md when checking requirements
4. Proposal structure (mapped to 様式1 sections; strictly observe character limits):
   • I. 研究目的 (80–400 characters) — Purpose of the research, target phenomena/challenges
   • II. 研究方法 (160–800 characters) — AI application per work phase, data, evaluation metrics, validation methods
   • III. AI利活用の妥当性・実現可能性 (160–800 characters) — Limitations of conventional methods, significance of AI adoption
   • IV. 達成目標 (100–500 characters) — マイルストーン at mid-term / 中間（3ヶ月後）and final / 最終（6ヶ月後）
   • V. AI利活用のノウハウ抽出・共有の実現計画 (60–300 characters) — Community sharing and cross-domain expansion
     • Reuse assets/knowhow-sharing-template.md when producing the know-how sharing section
   • Reference: Figure attachment (max 1 figure) — Select from Phase 1b draw.io architecture diagrams
   • VI. 研究業績等 (max 5 items) — Bullet list of journal papers, conference presentations, and books
   • Research infrastructure plan (Azure architecture)
   • Expense plan and 積算根拠 / cost justification (直接経費 ≤ ¥5M)
5. Generate proposal: Save as output/{project-name}/phase3-proposal.md
   • Reuse assets/proposal-template.md when producing the proposal
6. Quality review: Verify alignment with 公募要領 requirements and logical consistency

Deliverables

• output/{project-name}/phase3-proposal.md: SPReAD proposal (complete version)

Quality Gates

• All mandatory items in the 公募要領 are covered
• Innovativeness as "AI for Science" is clearly articulated
• Research plan and Azure architecture are consistent
• Expense plan justification matches the cost estimate document
• Research schedule includes milestones
• The Japanese text reads naturally and clearly

Gotchas

• 🚫 Do NOT fill expense figures using LLM training data or estimated values. Cost justification must be based on Azure Retail Prices API-retrieved unit prices from output/{project-name}/phase2-cost-estimate.md. If Phase 2 does not exist, run spread1000-cost-estimator first
• SPReAD's core theme is "innovating scientific research through AI." Emphasize research approaches that are only possible with AI, rather than simple replacements of existing methods
• Adjust the cost estimate so the expense plan stays within the SPReAD budget ceiling (直接経費 ≤ ¥5M)
• Do not forget to address research ethics and data management policies
  • Read references/dmp-guide.md when producing data management sections
• Strictly observe the character limits for each section (Japanese/English character ranges are specified per section)
• 様式1 must be submitted as Excel (PDF conversion is not allowed). Count characters in the template and keep within limits
• Maximum 1 figure. Select the most appropriate diagram from the Phase 1b draw.io architecture diagrams

Security Guardrails

• Do not include actual patient data, personally identifiable information, or confidential data in the proposal. Data examples must be anonymized or use dummy data
• References to IRB (research ethics review) and DMP (data management plan) are mandatory
• For research handling medical data, explicitly state compliance with data protection regulations (個人情報保護法, HIPAA, etc.)

Validation Loop

1. Generate the proposal
2. Check:
   • Are all 公募要領 items covered?
   • Is the proposal consistent with Phase 0–2 deliverables?
   • Is the text logically coherent?
3. If any check fails:
   • Add missing items
   • Fix inconsistencies
   • Refine the text
4. Finalize the deliverable only after all gates pass