oss-launch

Take a repo open source the right way. Run /oss-launch and it scans your repo, asks only what it cannot infer, then generates a tailored open-source file collection: README, LICENSE, CONTRIBUTING, CODE_OF_CONDUCT, SECURITY, CHANGELOG, .gitignore, GitHub issue and PR templates, and CI. Then it helps you launch: badges, metadata, releases, demo GIFs, and a Show HN / Reddit / YouTube playbook.

A stylized walkthrough of the workflow, not a screen recording (higher-res clip). For a real run's actual output, see example/: a real before/after scaffold with real audit.sh scores (1/16 -> 16/16), not hand-written boilerplate.

If this is useful, star it. It helps other solo maintainers find it.

Why

Most solo projects never get the boring-but-decisive OSS files: a real LICENSE, a contributing guide, a security policy, a README that pitches in one line. oss-launch does the scan and the scaffolding in one pass, adapted to your stack and repo, instead of pasting boilerplate that still says {{REPO}}.

Quick start

Prerequisite: shell/bash execution, to run scripts/*.sh. Every entry point here is a bash script, so on Windows use Git Bash or WSL; cmd.exe and PowerShell cannot run them.

This is a skill for agent harnesses. For Claude Code, install it where the harness discovers skills:

git clone https://github.com/AnayDhawan/oss-launch.git ~/.claude/skills/oss-launch

Then, inside any repository you want to open source:

/oss-launch

On other harnesses, bash scripts/build-agent-dirs.sh emits ready-to-copy bundles for Codex CLI, Cursor, Gemini CLI, Windsurf, and Aider, each built to that platform's documented format and verified self-contained by tests/verify-bundles.sh in CI. Claude Code is the one this is dogfooded on daily; none of the others has been live-trigger-tested in a real install, and Cursor is not planned for it (its CLI has no headless mode to drive). See AGENTS.md for the per-harness table of what is and isn't verified (details in tests/agent-bundle-verification.md), plus a shell-only path that works on any agent.

It will scan, report gaps, ask a few questions, and generate the files. The shell helpers also run standalone:

bash ~/.claude/skills/oss-launch/scripts/audit.sh .      # gap checklist for the current repo

No agent? Run it headless

scripts/apply.sh runs the same scan -> generate -> re-audit flow from a config file instead of an agent Q&A round, so it works in plain CI or a terminal with no agent loop:

cp templates/oss-launch.config.example oss-launch.config   # fill in AUTHOR, SECURITY_EMAIL, TAGLINE
bash scripts/apply.sh /path/to/your/repo --config oss-launch.config

It never overwrites an existing file, and skips README.md generation (prose is an agent-only step) rather than emitting boilerplate. See AGENTS.md for per-harness agent setup instead.

What it does

/oss-launch
  0. Scan      stack, existing files, git remote, secrets/brand leaks
  1. Report    gap table (scripts/audit.sh)
  2. Ask       license, author, security contact, type, tagline (only what is unknown)
  3. Generate  the OSS file collection from templates/, adapted and placeholder-filled
  4. Re-audit  what was created/updated, what is still manual
  5. Launch    metadata, CI, releases, demo GIF, Show HN / Reddit / YouTube (on demand)

What is in here