Grant proposal drafter that maps the funder's criteria to your evidence before writing a word
A Claude Project for nonprofits, researchers and small teams applying for grants: upload the call for proposals and your organization's materials, and it builds a criteria compliance matrix, finds evidence gaps, then drafts each section within the word limits.
Agent in Writing by AISkillHub Library
What it does
Grant applications are scored against criteria, but drafts are often written from the applicant's point of view and miss what reviewers look for. This project starts with the funder's call, builds a matrix of every requirement and scoring criterion, and only then drafts sections that answer them with your evidence.
**What it does**
1. **Reads the call** (guidelines, scoring rubric, eligibility, required sections, word or character limits, deadlines) from project knowledge. 2. **Builds a compliance matrix:** every requirement and criterion, where it will be addressed, and the evidence you have for it. 3. **Lists gaps:** criteria with weak or no evidence, eligibility questions, and attachments you still need. 4. **Drafts sections one at a time** in the funder's order and terminology, within limits, using only facts from your materials: need statement, objectives, activities, outcomes and evaluation, timeline, organizational capacity, sustainability, budget narrative. 5. **Self-checks** each section against the rubric and reports the word count.
**When to use it**
- Foundation, government and research grants with a written call - Turning a previous proposal into a new application for a different funder - Reviewing a colleague's draft against the rubric
**What it will not do**
- It will not invent outcomes, statistics, partners, letters of support or past results. Missing evidence goes on the gap list with a suggestion of what kind of evidence would fit. - It does not confirm eligibility. It points out eligibility conditions for you to check with the funder. - It does not build the budget spreadsheet. It writes the budget narrative from figures you provide and checks that the narrative and numbers agree.
**Example**
For a call that scores "Evidence of community need (20 points)", the matrix shows your needs assessment survey as the evidence source, notes that it is from 2021, and the gap list asks whether more recent data exists. The need statement then quotes the survey figures with the year, rather than presenting them as current.
How to use it
**Set up as a Claude Project** (Projects are available on claude.ai; limits depend on your plan).
1. In Claude, open **Projects** and create a new project named `Grant: <funder name>`. 2. Open the project's instructions setting (the panel for custom instructions that apply to every chat in the project) and paste the full contents of `assistants/grant-proposal-drafter/instructions.md`. 3. Replace the variables in the pasted text: `{{organization_name}}`. 4. Add reference material to the project's knowledge area (upload files or paste text): the funder's call for proposals and scoring rubric, your organization overview, program description, past reports or evaluations, budget figures, and staff bios. Fill in `assistants/grant-proposal-drafter/org-facts-template.md` if your facts are scattered. Create one project per application so materials from different funders do not mix. 5. Start a new chat inside the project. Every chat there uses these instructions and can read the project knowledge.
**Using it:** start with `Build the matrix`. Review the gap list and paste answers. Then ask `Draft section: <name>` one section at a time. Ask `Check against rubric` on any section, including ones you wrote yourself.
Files
Save each file at the path shown, relative to your project root.