More screenshots
Works with the AI Agents you already use
Launch Check: Is Your AI-Built App Safe to Launch?
Turn your app's URL into a one-page launch report card: leaked keys, open Supabase tables, phone layout, Google and AI search, speed. Plain English, with a fix list to paste to your AI.
Free
About this skill
You built your app with AI and it works. But is your Stripe key sitting in the page code? Can strangers read your Supabase tables with the key every visitor downloads? This is the most common real failure: a researcher found 170 of 1,645 Lovable showcase apps whose database endpoints could be read or written with the public key (CVE-2025-48757). On our seeded demo app, Launch Check found every planted problem in 7 seconds and the score went 30 -> 93 after the fix list was applied.
One command, one report card:
- Security (passive): secret keys in the page and its JavaScript (Stripe live/restricted, OpenAI, Anthropic, AWS, GitHub, Slack, SendGrid, private keys, Supabase service_role; the normal Supabase anon key is not flagged as a leak), security headers, HTTPS redirect, cookie flags, mixed content. Values are cut to their first 4 characters.
- Open database test (
--owner-check, your own site only): finds the Supabase URL and public key in your page and lists which tables anyone can read (GET only, 1 row per table, contents never shown). Firebase Realtime Database too. - Phone design at 375 px, found by Google & AI search, speed & basics (page weight, JavaScript errors, broken links, title, favicon).
- 0-100 score, traffic lights, plain English ("Anyone can see your Stripe secret key in your website's code"), a one-page HTML report and a fix list to paste to your AI.
--repo .also finds.envfiles committed to git.
Measured (2026-10-21, commands in the bundle)
| Site | Time | Score | Verified finding |
|---|---|---|---|
| Seeded demo app (before) | 7 s | 30, 8 high | Stripe key shape + service_role key in the bundle, 2 open tables, phone overflow |
| Same app after the fix list | 7 s | 93 | 0 high, 0 security findings |
| bolt.new (live, passive) | 49 s | 85 | low-contrast text on phone, 1.4 MB of JavaScript |
| v0.app (live, passive) | 48 s | 86 | 4.7 MB homepage, 78 tap targets too small |
False positives we found on live sites and fixed: rate-limited links (HTTP 429) counted as broken, analytics cookies asked for HttpOnly.
What it does not do: it is not a penetration test. No login or paywall testing, no API fuzzing, no write tests, scripts on other hosts (CDNs) are not scanned. Some sites block automated visitors (HTTP 403): run it on your preview URL.
Works in Claude Code, Cursor, Codex CLI, Copilot and any SKILL.md agent.
Privacy: readable Python, standard library (Playwright optional). No telemetry, no account. It downloads only what a visitor's browser downloads and never sends a found secret anywhere.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 26 days ago
- Free with an Agensi account
Requires
Permissions
Allowed Hosts