Skip to content
Mirayaمرايا
All projects
ToolingBetaOpen source

launch-kit

An open-source Claude Code skill that reads your repo and prepares an honest launch pack: positioning, GitHub polish, channel drafts and a 14-day plan.

From the README

launch-kit

Prepare an honest launch for any open-source repo.

launch-kit is a Claude Code skill that reads your repository and writes a launch pack to launch/: positioning, GitHub polish with concrete README edits, ready-to-edit drafts for Show HN, Reddit, Product Hunt, X, LinkedIn and dev.to, and a 14-day launch calendar. Everything stays on your machine as markdown. You review it, rewrite it in your own words and post it yourself.

test release License: MIT Claude Code plugin

launch-kit: prepare an honest launch for any open-source repo

Quick start

In Claude Code:

/plugin marketplace add miraya-dev/launch-kit
/plugin install launch-kit@launch-kit

Then open the repository you want to launch and ask:

Prepare a launch pack for this repo.

or call the skill directly: /launch-kit:launch-kit. The pack appears in launch/.

What you get

FileWhat's inside
00-facts.mdFacts collected by a script: license, languages, tests, CI, demo media, community files, README outline
01-positioning.mdOne-liner, who it's for, 3 key benefits with proof from the repo, an honest comparison with alternatives
02-github-polish.mdRepo description, 5 to 10 topics, a social preview image brief, a README checklist with ready-to-paste edits, LICENSE / CONTRIBUTING / good-first-issues check
03-show-hn.mdTitle options and notes for your Show HN text, plus likely questions with factual answers
04-reddit.mdSuggested subreddits with the rule to check in each, and one tailored post per subreddit
05-product-hunt.mdTagline, description, maker's first comment, thumbnail and gallery brief
06-x-thread.mdAn X thread, every post counted the way X counts
07-linkedin.mdA LinkedIn post with an opening that fits before "…see more"
08-article-outline.mdA dev.to / Hashnode article outline with front matter
09-calendar.mdA 14-day launch calendar and a post-launch follow-up checklist
README.mdThe index: top 3 fixes before launch and the placeholders left to fill

See a real pack: examples/launch-kit/ is the pack launch-kit wrote for its own launch, unedited (how it was made). Three excerpts:

From 02-github-polish.md, the README checklist:

ItemStatusConcrete edit
Demo GIF or screenshot in the first screenpartialLine 11 shows docs/social-preview.png, a static card. Nothing shows the skill running. Record a 15 to 20 s GIF (type the prompt, files appear in launch/, check-limits.mjs prints "0 over the limit"), save it as docs/demo.gif and put it right under the badges.

From 05-product-hunt.md, the tagline (51/60):

Turn your repo into a launch pack you post yourself

From 06-x-thread.md, the first post (256/280, counted the way X counts):

I built launch-kit, an open-source Claude Code skill that reads your repo and writes a launch pack: positioning, README fixes, drafts for Show HN, Reddit, Product Hunt, X, LinkedIn and dev.to, and a 14-day plan.

You edit the drafts and post them yourself.

Why

Building the project is one job; launching it is another. Most makers either skip the launch or paste the same post everywhere, and some communities remove it for breaking their rules. launch-kit does the preparation from what is actually in your repo: it finds the gaps visitors will notice first, writes drafts tailored to each channel, and keeps you within each community's rules.

How it works

  1. scripts/repo-facts.mjs collects facts deterministically (no network, no dependencies): name, languages, license, tests, CI, demo media, community files, release tags.
  2. Claude reads the README, manifests, main source files and docs, and looks for honest alternatives.
  3. For each file of the pack it reads one reference (the channel's rules and best practices) and fills in a template.
  4. scripts/check-limits.mjs checks every length-limited field: the Product Hunt tagline (60), Hacker News titles (80), X posts (280, with links counted as 23 and emoji as 2), the LinkedIn opening, GitHub topics.

The skill uses progressive disclosure: a short SKILL.md, one reference per channel loaded only when needed, and templates for every file of the pack.

Install

As a Claude Code plugin (recommended)

/plugin marketplace add miraya-dev/launch-kit
/plugin install launch-kit@launch-kit

From a terminal: claude plugin marketplace add miraya-dev/launch-kit then claude plugin install launch-kit@launch-kit. Update later with claude plugin update launch-kit@launch-kit.

Manual install

Copy only the skill folder into your personal skills folder:

git clone https://github.com/miraya-dev/launch-kit
mkdir -p ~/.claude/skills
cp -r launch-kit/skills/launch-kit ~/.claude/skills/

On Windows (PowerShell):

git clone https://github.com/miraya-dev/launch-kit
New-Item -ItemType Directory -Force "$HOME\.claude\skills" | Out-Null
Copy-Item -Recurse launch-kit\skills\launch-kit "$HOME\.claude\skills\"

Installed this way the skill is called /launch-kit. For one project only, copy it to .claude/skills/launch-kit/ in that project instead.

Requirements: Claude Code. Node.js 18 or newer for the two helper scripts; without Node, Claude collects the facts by hand.

Usage

Ask in plain words, or pass arguments to the skill:

Prepare a launch pack for this repo.
/launch-kit:launch-kit ../my-project 2026-11-17
Prepare the launch of this repo. I'm Sam (@sam on X). Launch day is Tuesday 17 November, I'm in Berlin. Skip Product Hunt and LinkedIn.
Only do the GitHub polish and the Show HN notes.
Here are two posts I wrote before. Match my voice in the drafts: …

The skill only writes inside launch/. When you're ready, ask: "Apply the README edits from launch/02-github-polish.md."

The scripts also work on their own:

node skills/launch-kit/scripts/repo-facts.mjs path/to/repo --format md --out path/to/repo/launch/00-facts.md
node skills/launch-kit/scripts/check-limits.mjs path/to/repo/launch

Ethics

These rules are written into SKILL.md as hard rules. The skill declines requests that break them and offers the honest alternative.

  • No fake engagement: no bought stars, votes or followers, no bots, no second accounts, no AI-written comments posing as users.
  • No star-for-star, vote rings or upvote groups. Asking people to try it and give feedback is fine; asking them to vote is not.
  • No astroturfing: no posing as a happy user, no seeded questions, no undisclosed paid posts.
  • Say you're the maker in every post.
  • Follow each community's rules. Every channel file starts with a link to its rules; if they say no, skip that channel.
  • Only true claims: nothing invented, and no running down alternatives.
  • Drafts are drafts. Hacker News asks for no generated text in posts or comments, so for HN the pack gives notes and facts, and you write the words.

The skill never posts, votes, comments or signs up anywhere.

FAQ

Does it post anything for me? No. It writes markdown files in launch/ and never posts, votes, comments or signs up anywhere. You rewrite the drafts and post them yourself.

Can I see what it does without Claude Code? Yes. examples/launch-kit/ is the pack it wrote for its own launch. The two scripts also run on their own with Node.js 18+: node skills/launch-kit/scripts/repo-facts.mjs path/to/repo --format md prints a launch-readiness checklist for any repo. The drafts themselves need Claude Code.

Do the scripts send my code anywhere? No. repo-facts.mjs and check-limits.mjs only read local files: no network, no dependencies, and repo-facts.mjs never runs git. Claude reads your files the way it does in any Claude Code session.

Which languages and ecosystems does it understand? Claude reads code in any language. repo-facts.mjs also reads the manifests of 13 ecosystems (npm, PyPI, crates.io, Go modules, Packagist, pub.dev, JSR, RubyGems, Maven, Gradle, NuGet, Swift Package Manager, Hex) and detects 10 CI providers.

Why does the Hacker News file have notes instead of a post? Hacker News asks for no generated text in posts or comments, so the pack gives you facts and likely questions, and you write the words.

Limitations

  • You still do the launch. The pack saves hours of preparation, but you rewrite, post, and stay to answer comments.
  • Rules change. The references were checked in October 2026. Always read each community's current rules; subreddit suggestions are starting points, and the skill can't see current sidebars.
  • Comparisons can be out of date. Claude uses what it knows, plus web search when available. Check them before posting.
  • Facts come from files only. repo-facts.mjs doesn't know your stars, traffic or GitHub releases (only local tags). It understands common .gitignore patterns, but not negations.
  • No images. It writes briefs for the social preview and Product Hunt gallery; you or a designer make them.
  • English by default. Ask for another language if your community writes in one.

Contributing

Issues and pull requests are welcome. See CONTRIBUTING.md and the code of conduct. Run the tests with npm test (Node.js 18+, no dependencies). Changes are listed in the CHANGELOG.

License

MIT © 2026 Moncef El hirech (Miraya)


Built by Miraya, a software studio by Moncef El hirech.

Releases