Launch Audit

Finds what makes a launch look broken, fixes it in your code, then plans launch day.

FREE
BETTER WITH THE TAILKITS MCP
MIT
4 FILES · 17 KB
V1.0.0

INSTALL

One command, any agent

$ mkdir -p ~/.claude/skills && curl -fsSL https://tailkits.com/skills/tailkits-launch-audit.tar.gz | tar -xz -C ~/.claude/skills

Installs into ~/.claude/skills. For one project only, use .claude/skills instead. Then start a new Claude Code session and ask. Claude Code skills docs

$ mkdir -p ~/.agents/skills && curl -fsSL https://tailkits.com/skills/tailkits-launch-audit.tar.gz | tar -xz -C ~/.agents/skills

Installs into ~/.agents/skills. For one project only, use .agents/skills instead. Codex reads skills from ~/.agents/skills. Start a new session and ask. Codex skills docs

$ mkdir -p ~/.gemini/skills && curl -fsSL https://tailkits.com/skills/tailkits-launch-audit.tar.gz | tar -xz -C ~/.gemini/skills

Installs into ~/.gemini/skills. For one project only, use .gemini/skills instead. Project skills load only in a trusted folder. Start a new session and ask. Gemini CLI skills docs

$ mkdir -p ~/.cursor/skills && curl -fsSL https://tailkits.com/skills/tailkits-launch-audit.tar.gz | tar -xz -C ~/.cursor/skills

Installs into ~/.cursor/skills. For one project only, use .cursor/skills instead. Cursor also reads ~/.claude/skills, so one install can serve both. Cursor skills docs

$ mkdir -p ~/.config/opencode/skills && curl -fsSL https://tailkits.com/skills/tailkits-launch-audit.tar.gz | tar -xz -C ~/.config/opencode/skills

Installs into ~/.config/opencode/skills. For one project only, use .opencode/skills instead. OpenCode also reads ~/.claude/skills and ~/.agents/skills. OpenCode skills docs

$ mkdir -p ~/.grok/skills && curl -fsSL https://tailkits.com/skills/tailkits-launch-audit.tar.gz | tar -xz -C ~/.grok/skills

Installs into ~/.grok/skills. For one project only, use .grok/skills instead. Grok also picks up skills installed for Claude Code. Grok skills docs

Prefer to read it first? Download tailkits-launch-audit.tar.gz and open it.

THEN ASK

Is https://acme.com ready to launch?
Why does my link preview look blank?
Audit localhost:3000 before we ship

SKILL.MD

What the agent reads

tailkits-launch-audit/SKILL.mdMIT license

name: tailkits-launch-audit
description: Audits a landing page before launch (SEO head, Open Graph image, icons, headings, alt text, weight, render-blocking scripts, a visible CTA), fixes what it finds in the code, then builds a launch checklist and a prefilled Tailkits submission link. Use when the user is about to launch, asks "is my landing page ready", wants a launch or SEO audit, asks why link previews look broken, or wants to know where to launch a product.
compatibility: The full audit, checklist and submit link use the free Tailkits MCP (https://tailkits.com/mcp). scripts/check-head.mjs (Node 18+, no dependencies) covers localhost, previews and built files, which the MCP cannot fetch.

Tailkits launch audit

Find what will make a launch look broken (no preview image, wrong title, noindex left on, no CTA above the fold), fix it in the code, and leave the user with a dated launch plan.

The MCP (free, no account)

Steps 1, 4 and 5 use the Tailkits MCP tools launch_audit_page, launch_checklist and launch_submit_link. If they are not available, tell the user once:

claude mcp add --transport http tailkits https://tailkits.com/mcp

(Other agents: https://tailkits.com/mcp.) Then continue with the script in step 1 and the checklist in references/launch-checklist.md.

Workflow

1. Audit

A public https URL and the MCP is connected: call launch_audit_page with url. It returns a score per area and fixes ranked by impact. Anonymous use allows 30 audits a day.
localhost, a preview behind auth, a built file, or no MCP: run the offline subset. The MCP only fetches public https hosts, so local pages always use this:

node <skill-dir>/scripts/check-head.mjs http://localhost:3000
node <skill-dir>/scripts/check-head.mjs dist/index.html --json

<skill-dir> is the folder that contains this SKILL.md.

2. Fix

Work through the findings from errors to warnings. For each one, find where the project sets it (layout component, <head>, framework metadata API such as Next.js metadata or Astro’s layout props) and change it there, not in built output. Typical fixes:

Title 30 to 60 characters, meta description 70 to 160, one H1 that says what the product is.
Absolute og:image URL to a 1200x630 PNG or JPEG. No image yet: use the tailkits-og-image skill if installed.
twitter:card = summary_large_image, canonical URL, lang, viewport.
noindex removed, robots.txt and sitemap.xml served.
Main CTA visible in the first section; images have alt; scripts in <head> get defer.

Do not rewrite marketing copy here beyond length fixes. If the user wants that, run tailkits-copy-pass.

3. Re-audit

Run the same audit again and show before and after counts. Anything left that the user must decide (for example the real domain for the canonical) goes in the report as a question.

4. Plan the launch

Call launch_checklist with stage = pre (the week before), day (launch day) or post (the week after); call all three when the user asks for a full plan. Turn it into dated steps from the user’s launch date. Without the MCP, use references/launch-checklist.md.

5. Submit to Tailkits

Call launch_submit_link with name, url and tagline. It returns a prefilled https://tailkits.com/submit URL. Give the user the link; it never submits anything, a human does.

Report

End with: the score or error/warning counts before and after, the files changed, the open questions, and the next dated launch step.

MORE SKILLS

Pair it with

All skills and the one-line install for all five