What you can do with it
- Build: turn a product brief into a landing page with a clear offer, working action, mobile layout, and basic search metadata.
- Research: find audience questions, search intent, and the formats of leading pages (guide, table, calculator, tool), then map useful pages. It labels missing keyword volume instead of making up numbers.
- Audit: check the message, every visible CTA, trust, footer links, accessibility, technical SEO, and AI readability. Get a short, prioritized list of fixes with evidence.
- Implement: fix selected issues in the existing site without replacing a design that already works.
- Launch: verify the deployed page, links, analytics when selected, sitemap, search submission where account access exists, and automatic IndexNow notices for eligible new or changed pages.
Launch Check is a set of instructions for your coding agent, not a hosted audit service or a one-command CLI. It uses the repository, browser, and connected accounts you give the agent. It separates what it checked from what it could not verify.
Try it in one prompt
Paste this into an agent chat after installation:
$launch-check audit https://whoownswhat.xyz/ for conversion and SEO.
Tell me what is clear, what is confusing, and which fixes matter first.
Check the visible CTAs, mobile layout, technical SEO, sitemap, and footer.
Show evidence and mark anything you could not verify. Do not change the site.
You can also ask it to build, research, implement, or launch. For a full release, name the repository, hosting project, and domain, and explicitly ask it to publish. See the complete usage guide for copyable prompts for every mode.
Install
Copy the whole skills/launch-check folder, including its references and assets, into your agent's skills directory:
Codex: ~/.agents/skills/launch-check/
Claude Code: ~/.claude/skills/launch-check/
Use your agent's skill selection syntax, such as $launch-check in Codex or /launch-check in Claude Code. Then paste one of the prompts above. The skill has no mandatory paid service or runtime dependency. Browser, repository, Search Console, and analytics access are needed only for checks or changes that use them.
Already installed? Say “Update launch-check from GitHub to the newest version” in your coding agent, or run the included updater. It checks the main branch, backs up older installed copies, and updates on --apply.
Built with Launch Check
These are projects from avo42. They show different kinds of pages and products the skill has helped build or check.
| Project | What it does |
|---|
| WhoOwnsWhat | Shared ledger for co-founders to split project costs, track asset ownership, and hand things over cleanly. My latest project, still in development. |
| MakeSquishy | Free tool for kids to design, print, and fold paper squishies. |
| PadelTune | Personal padel training plans for strength, mobility, and match prep. |
| ReadNext | Helps pick one book to read next from an overflowing want-to-read shelf. |
| PlateWeek | Turns a food video into recipes and a practical shopping list. |
| Armando | Client management playbook and follow-through for agencies and consultancies. |
| AI Shift Field Guide | Plain-language guide to AI shifts in software, with a checklist for testing product ideas. |
| AI Markets, explained | A practical reading of AI spending, adoption, and where new products may fit. |
How it handles SEO research
The default research path uses accessible, attributable sources. The agent checks intent and the actual search results, then turns promising topics into page briefs. Numeric volume or difficulty appears only when a connected source supplies it. DataForSEO can add live keyword and SERP data, but it is optional and may have its own cost; the skill does not pretend that it is connected by default.
A brief now separates why someone searched from what the leading pages let them do. The agent opens leading organic results, records each URL and rank, and reports the confirmed format mix alongside volume and organic keyword difficulty when those metrics are available. It then recommends a guide, comparison, calculator, tool, or no page based on the evidence and what the product can actually deliver. For example, a query about splitting startup costs might need an explanation and a calculator. If pages cannot be inspected, their formats stay unverified and the recommendation is a hypothesis.
Mock layout only: the pictured formats were invented to show the report shape. A real run must inspect and cite the ranked pages.
For search discovery, the launch workflow checks canonicals, robots.txt, generated sitemaps, and Google/Bing submission when the right accounts are available. IndexNow can notify participating engines about page changes. None of these steps guarantees indexing or ranking.
More detail
I have been iterating on the public repo since late September 2026. Suggestions and real-world edge cases are welcome in GitHub Issues.
