Impact to Funding
For builders. Experimental v0, open source.
Put your skills and experience towards real problems, and build a credible case for support.
ItF is a skill for the AI agent you already use and are signed in to. Start with your situation: you don't need a theory of change, a project, evidence or a fundraising goal. If you're worried about work, you don't need to become a founder or work for free.
Install
Pick your agent, copy the command, and run it in your terminal. Conversation needs no signup, Node.js, MCP server or plugin. Your agent's model costs and privacy terms still apply.
Claude Code
Run this in your terminal, from the project where you'll use ItF. You need Git.
target='.claude/skills/impact-to-funding'
revision='4566892ad990131fb36f760e14c73e5f275600cc'
if [ -e "$target" ] || [ -L "$target" ]; then
echo "Destination already exists; stop and inspect: $target"
else
git clone https://github.com/zhiganov/impact-to-funding.git "$target" &&
git -C "$target" checkout --detach "$revision" &&
git -C "$target" rev-parse HEAD
fiOpen Claude Code in the same project and type /impact-to-funding. If the session was already open, run /reload-skills first.
Codex CLI
Run this in your terminal, from the project where you'll use ItF. You need Git.
target='.agents/skills/impact-to-funding'
revision='4566892ad990131fb36f760e14c73e5f275600cc'
if [ -e "$target" ] || [ -L "$target" ]; then
echo "Destination already exists; stop and inspect: $target"
else
git clone https://github.com/zhiganov/impact-to-funding.git "$target" &&
git -C "$target" checkout --detach "$revision" &&
git -C "$target" rev-parse HEAD
fiOpen Codex in the same project and type $impact-to-funding. You can also pick it from /skills.
OpenCode
Run this in your terminal, from the project where you'll use ItF. You need Git.
git clone https://github.com/zhiganov/impact-to-funding.git .opencode/skills/impact-to-funding
git -C .opencode/skills/impact-to-funding checkout --detach fbb8993060254ef63f5157d3b9ddcbb8bb4cf19bIn OpenCode, in the same project, type @impact-to-funding. This is an older snapshot, the one tested in OpenCode. It lacks some newer guidance. Use an empty destination.
Installs into this project only and changes no global settings. If something already exists at the destination, it stops instead of overwriting. Full guide, personal installs and updates: installation guide.
Claude Code and Codex CLI were each checked with one non-interactive run. OpenCode was install-tested with an older snapshot. Menus, desktop apps and usefulness were not tested.
MCP server
An optional, read-only MCP server over ItF's reference library.
It lets any MCP client search the library's cases, candidate patterns and lenses with three tools: search_references, get_reference and library_info. It runs locally, reads a snapshot you prepare from your installed copy, and needs no account or API key. You don't need it to use the skill.
It needs Node.js 22 or later, npm and Git. Prepare the snapshot once, from your install folder:
cd .claude/skills/impact-to-funding # your ItF install folder
npm ci --ignore-scripts --prefix mcp
mkdir -p outputs
node mcp/prepare.mjs --revision "$(git rev-parse HEAD)" --output outputs/library-snapshot.jsonThen add a stdio server to your agent's MCP settings: command node, arguments <install folder>/mcp/server.mjs --snapshot <install folder>/outputs/library-snapshot.json, using absolute paths. Details in the MCP guide.
Three ways to start
After you invoke the skill, try one of these. They're starting points, not commands. A few sentences about your situation are enough.
Find a useful contribution
Help me explore where my skills and available time could be useful, including contributing to existing work. Use what I supply, mark assumptions, and keep this private. Do not contact anyone.
Develop shared work and learn
Help me describe my role in this shared project, check it against my capacity, and examine what we know about its effects. Preserve other contributors and disagreement. Do not collect or publish anything.
Explain what support would help
Help me prepare a private case for sustaining this work. Separate observations from proposals, use only known costs and commitments, and keep uncertainty visible. Do not invent impact, submit an application or contact a funder.
A useful result might be a possible contribution, a clearer question, a smaller commitment or a decision to stop. It doesn't have to be a report.
Limits
- Private by default. A local file can still go to a remote model when your agent reads it. Use only material you may share with that provider.
- No action without your say-so. A draft doesn't authorise publishing, contacting people, applying or paying.
- No certification or ranking. Outputs and completed assessments don't prove impact.
- Limited evidence. Development trials were small and the examples are fiction. We haven't verified usefulness to people or fundraising results.
Original writing is CC BY-SA 4.0; code is MIT. ItF draws on public Moral Ambition principles and isn't affiliated with or endorsed by the School for Moral Ambition or Rutger Bregman.