Generate evidence-backed OSS maintainer briefs and application drafts from public GitHub activity.
oss-grant-kit helps maintainers collect the boring-but-important evidence needed for OpenAI/Claude OSS programs, grants, sponsorships, and maintainer support applications.
It does not fabricate impact. It scans public GitHub data, ranks candidate repositories, highlights gaps, and writes draft application text that the maintainer should review before submitting.
Small maintainers often have real work scattered across many repositories, but application forms ask for a clean story:
- Which project should I lead with?
- What public evidence supports it?
- What are the weak spots?
- How do I describe my maintenance plan honestly?
This tool turns a GitHub profile into a verifiable maintainer packet.
npx oss-grant-kit brief --github hatrd --target all --out ./oss-briefOr from a local checkout:
git clone https://github.com/hatrd/oss-grant-kit
cd oss-grant-kit
node src/cli.js brief --github hatrd --target all --out ./demo-outOutputs:
brief.md— maintainer brief and top repository rankingevidence.json— raw facts used by the generatorapplication-openai.md— OpenAI Codex for OSS draftapplication-claude.md— Claude for Open Source draft
oss-grant-kit brief --github <username> [options]Options:
--github <user>: GitHub username to scan--repos <list>: optional comma-separated repo allowlist--target <target>:all,openai-codex-oss, orclaude-oss--out <path>: output directory or markdown file--limit <n>: number of repositories to enrich with extra API calls--token <token>: optional GitHub token;GITHUB_TOKENis also supported--include-external false: skip GitHub cross-repository contribution search--include-forks false: scan only repositories directly owned by the account. By default the tool now scans all visible user repositories, including forks, because maintained forks and community handoffs can be important evidence.--discover-platforms false: disable automatic README/homepage discovery for external platform links--platform <spec>: add non-GitHub user adoption evidence. Supported specs includesteam:<publishedfileid>,openvsx:<namespace.extension>, andvscode:<publisher.extension>. Repeat the flag or pass comma-separated specs.
Example for a Don't Starve Together mod on Steam Workshop:
oss-grant-kit brief --github hatrd --platform steam:1234567890 --out ./oss-briefThe current heuristic considers:
- stars and forks
- recent commits
- releases
- README and license presence
- original repo vs fork
- relevant keywords such as AI, agent, devtool, CLI, security, automation
- visible maintenance gaps
The score is a sorting aid, not a truth oracle.
- External contribution detection: public events plus GitHub search for authored commits and involved issues/PRs outside repositories owned by the scanned account.
- External platform usage: Steam Workshop, Open VSX, VS Code Marketplace, and other non-GitHub user platforms can be attached or discovered as adoption evidence, so projects like DST mods and editor extensions are not undervalued just because their users are outside GitHub.
--include-external falseto skip cross-repository contribution search.
- Generated applications are drafts.
- Do not submit without human review.
- Do not inflate metrics or imply maintainership you do not have.
- External form submission is intentionally out of scope.
- richer PR/review activity analysis
- package download enrichment for npm/PyPI/crates.io
- richer external platform connectors: Steam Workshop author pages, VS Code Marketplace, browser extension stores, mod portals
- GitHub Action wrapper
- project-specific templates for grants and sponsorships
- OpenClaw Skill wrapper for interactive workflows
MIT