“Can we ship?”
Let your AI agent answer with evidence.
9 Agent Skills + 1 subagent that run a real pre-release QA pass (smoke tests, visual regression, accessibility, performance budget, changelog sanity) and produce one GO / NO-GO report. For Claude Code, Cursor, Codex and other agents that support Agent Skills.
Advice is free. Verdicts need procedures.
Your agent says “all tests pass” and you ship. Then the pricing page throws a JS error on mobile, an image has no alt text, and the bundle grew 40%. Release Gate Kit gives your agent a step-by-step procedure and the scripts to back it up: it runs the checks, reads the evidence, and reports what passed, what failed, and what it could not verify.
Runs, not just reads
5 zero-dependency Node.js scripts plus Playwright and axe-core specs. Every check ends in an exit code and a JSON result.
Honest by design
Anything that could not be verified is reported as NOT RUN, never as passed.
Safe by default
The skills never deploy, tag, publish, push, or overwrite screenshot baselines on their own.
What's included
Plain folders you copy into your project. Each one has a SKILL.md with trigger conditions, steps, and output rules.
release-gate
Runs build → smoke → visual → a11y → perf → changelog and writes RELEASE_REPORT.md with GO / GO WITH WARNINGS / NO-GO.
qa-setup
Detects Next.js, Vite, Astro, SvelteKit, CRA or a static site and installs a Playwright config that tests the production build.
smoke-tests
Checks every key route for bad status codes, uncaught exceptions, console errors, failed assets and missing content, on desktop and mobile.
visual-regression
Full-page screenshots, masking of dynamic areas, a Linux baseline workflow, and rules for judging diffs.
a11y-audit
axe-core WCAG 2.2 AA checks. Fails on critical or serious issues, suggests fixes for your framework, and adds a manual checklist.
perf-budget
Gzipped JS/CSS budgets, your largest files, and growth since the last release. Lighthouse is optional.
changelog-sanity
Checks the semver bump against the last tag, the changelog entry, major bumps for breaking changes, and leftover .only / debugger.
flaky-test-triage
Reruns each test N times, reports its pass rate, and gives a root-cause table and fix-or-quarantine rules.
ci-release-gate
A GitHub Actions workflow with browser caching, artifacts and a job summary, plus a GitLab CI snippet.
qa-gatekeeper (subagent)
A Claude Code subagent that runs the gate in its own context and returns a short verdict.
Sample output
A real run on a small test site with two planted bugs.
$ node .claude/skills/release-gate/scripts/release-gate.mjs ▶ build … PASS build succeeded ▶ smoke … FAIL 1 passed, 1 failed, 0 flaky ▶ visual … PASS 2 passed, 0 failed ▶ a11y … FAIL 1 passed, 1 failed ▶ perf … PASS JS 0.1 KB gz, CSS 0.1 KB gz ▶ changelog … PASS version 1.1.0 (last tag v1.0.0) 🔴 NO-GO – report: qa-reports/RELEASE_REPORT.md # qa-reports/RELEASE_REPORT.md ## Blockers ### smoke - about loads cleanly [desktop]: HTTP 404: /x.png ### a11y - about [desktop]: [critical] image-alt: Images must have alternative text (1 node) # after fixing both issues 🟢 GO – report: qa-reports/RELEASE_REPORT.md
Works with your agent
Any tool that reads the open Agent Skills format (folders with a SKILL.md).
Requirements: Node.js 18+ and a web front end. The qa-setup skill installs @playwright/test and @axe-core/playwright for you. Release Gate Kit is a set of files you install; it is not a hosted AI service and has no model access built in.
Release Gate Kit v1
- 9 skills + 1 subagent, scripts, templates and CI workflow
- Instant ZIP download after checkout
- Single-user commercial licence: use it in unlimited projects
- Free 1.x updates
- 30-day money-back guarantee
Secure checkout by Creem, our merchant of record. Taxes (e.g. VAT) may be added at checkout depending on your location.
FAQ
Is this just Markdown prompts?
No. Each skill is a step-by-step procedure, and the bundled scripts actually run. The gate was tested on a sample site with planted bugs: it returned NO-GO for a 404 image request and a missing alt attribute, then GO once both were fixed.
Free QA skills exist. Why pay?
Free collections are broad libraries of guides. This pack is one opinionated workflow that ends in a verdict, with working scripts, CI and report formats that fit together.
What do I receive after purchase?
A ZIP file (about 40 KB) with the skills, subagent, README with install steps, changelog and licence. The download link is shown after checkout and sent to your email.
Backend, mobile apps, Python?
v1 focuses on web front ends. The changelog and flaky-triage skills work in any Node.js repo.
Can I use it at work or for clients?
Yes, in unlimited projects. One licence per person; you may not resell or redistribute the pack.
Does it include an AI model or API access?
No. You use your own agent (Claude Code, Cursor, Codex…). The kit adds instructions and scripts to it.
Refunds?
30 days, no questions asked. See the refund policy.
Contact & support
Questions before or after buying, licence questions, or bug reports: email support@factppt.com. We reply within 2 business days.
Seller: YAN XIONG, an individual sole proprietor based in Kunming, China, China. Orders are processed by Creem, our merchant of record, which handles payment, taxes, invoices and refunds.