grant-proposal

What it does

Drafts a grant proposal from validated research ideas, guiding the user through a multi-phase workflow to produce a reviewer-ready document tailored to a specified grant type and sub-type. It supports multiple funders (KAKENHI, NSF, NSFC, ERC, DFG, SNSF, ARC, NWO, GENERIC) and handles language and formatting requirements. The process includes literature grounding, gap verification, aims design, a claims-evidence matrix, narrative arc planning, timeline milestones, and a structure review before drafting sections.

How it works

• Phase 0 parses the input arguments to identify the research direction, grant type and sub-type, and any overrides (format, language, review rounds). It also collects context from the project directory (IDEA_REPORT.md, literature notes, publication lists, prior reviews, and a prior GRANT_STATE.json if present).
• Phase 1 conducts literature and landscape positioning by invoking /research-lit to ground the proposal, then searches for competing funded projects and runs /novelty-check to articulate a gap statement. It then pauses for user confirmation before proceeding unless AUTO_PROCEED is true. The phase ends with writing GRANT_STATE.json (phase: 1) and awaiting user input.
• Phase 2 designs Narrative Structure & Aims, including: defining 2–4 Specific Aims with deliverables, building a Claims-Aims-Evidence Matrix, crafting a narrative arc (Problem → Why Now → What We Propose → Why It Will Work → What We Will Deliver), and outlining a Timeline & Milestones. It ends with a checkpoint that presents the designed structure and seeks user feedback before drafting.
• Phase 3 (not fully shown here) would draft the proposal sections per the chosen grant type template, producing complete prose documents rather than outlines.

All outputs are formatted according to the specified OUTPUT_FORMAT (markdown or latex) and LANGUAGE, and use the prescribed directory grant-proposal/ for generated files. The system persists state in grant-proposal/GRANT_STATE.json after each phase and uses phase-based resume logic.

When to use it

Use when a user needs to turn research ideas into a fundable grant application and wants agency-specific structure and language, including KAKENHI, NSF, NSFC, ERC, DFG, SNSF, ARC, NWO, or GENERIC formats. Triggered by phrases like "write grant", "grant proposal", or equivalents in various languages.

What it can touch

• Commands and tools referenced: /research-lit, /novelty-check, /research-review. These are invoked as part of the workflow to ground the proposal and solicit expert feedback. The skill reads existing files like IDEA_REPORT.md, refine logs, and GRANT_STATE.json from the grant-proposal directory to resume context.

Caveats

• The workflow requires explicit user confirmation at designated checkpoints unless AUTO_PROCEED is set to true. The process is designed to pause for PI judgment at each stage. State persistence uses grant-proposal/GRANT_STATE.json and may store phase, grant_type, language, and a gap/aims snapshot. No outcomes are promised; the system only outlines structure, alignment, and drafting steps as described.