Proof leads the Crew. It writes Playwright tests that click through sign-up, login, checkout and cancel like a stranger would, signs a test Stripe webhook to prove payment grants access exactly once, and leaves a GitHub Actions workflow that reruns them on every pull request. On our demo app the scanner said 99/100. Proof found that nobody could sign up.
The money example: a customer cancels and keeps Pro, because the app never handles customer.subscription.deleted. Tally fixes that drift both ways, with signed, idempotent webhooks that write the subscription's real state. Backing it all: a scanner with the same findings every run, file and line on each, and an exit code that fails CI on anything high or critical.
shipready_scan.py, groups findings by ownership-report.md + Excel scorecard(select auth.uid())app_metadata or a locked table, never user_metadatastripe_events<img>playwright.config.ts + GitHub Actions workflow you can rerun on every PRUnzip the download, then in Claude Code: claude plugin install ship-ready-crew --marketplace ./ship-ready-crew. Others: drop the folder or paste the system prompt.
"Is this app safe to launch?" Harbor maps the app, scans, routes and reports, with every finding owned by one specialist.
Minimal diffs and SQL with rollbacks. You act at the gates. Re-scan until the verdict turns green.

49 findings · 13 critical · DO NOT SHIP
1 low, accepted · READY
Included: promptpad-app/ (broken), promptpad-app-fixed/, REPORT-before.md, REPORT-after.md, findings JSON, fixes.patch (~510 lines), filled Excel scorecard and a 6-step walkthrough, including what the example does not prove.
Yes. Built for people who shipped with Lovable, Bolt, v0 or Cursor and want to know it's safe. Harbor explains every finding in plain English, says why it matters and gives the exact fix. You make the final calls.
Claude Code installs it in one line as a plugin. The same skills work in Cursor, Codex and other AGENTS.md agents, and as uploadable skills in Claude.ai. ChatGPT, Claude Projects and Gemini get paste-in system prompts.
You download a ZIP right away. It installs in Claude Code with one line from the unzipped folder, and has every agent folder, per-agent ZIPs for Claude.ai, system prompts for chat apps, the scanner, templates and the worked example. Want updates through Claude Code? Send us your GitHub username and we add you to the private buyer repo.
The $39 Crew covers your own apps plus up to 3 client projects a year; studios doing more: contact us. Don't resell or republish the pack.
Yes, and you can check. The packs are plain Markdown skills plus one offline Python scanner you can read. There is no npm package and no install script, so nothing runs when you add it. Worth asking of any skills pack: on 22 Oct GitHub flagged npm versions of two popular skills collections as malware.
No. The crew works in your local repo and proposes minimal diffs. Two human gates are built in. Rotating keys, changing dashboards, applying production migrations and deploying are always your call.
A checklist tells you what to look for. The Crew does the looking: a deterministic scanner with file:line evidence, a lead agent that routes work to specialists, a score you can track, and a full worked audit so you can see exactly what you'll get.
The Crew ($39, one-time) and the Audit ($129 per app) will be available here as soon as the store listing is live. Lite is free now.