Skip to content

Autonomous development for real products. Governed agents build from specs you approve, follow the rules you set, and nothing merges until another agent proves it. Your decisions, judgement and taste are the bar that work must clear. Declare your vision and let agents build it.

3.6 KB
Raw Blame
{
  "blueprint": 1,
  "name": "web-app",
  "title": "Web app",
  "version": 1,
  "kind": "product",
  "by": "Pullboard",
  "summary": "For building a web app with AI agents. One agent builds each change, a second checks it against your rules before it reaches main, and only your decisions reach you.",
  "prefix": "W",
  "gate": "npm test",
  "rules": [
    { "key": "W1", "text": "Every change ships with a test that fails without it.", "why": "a test that passes either way proves nothing." },
    { "key": "W2", "text": "Main only moves when the whole test suite passes.", "why": "two changes can each pass alone and still break each other." },
    { "key": "W3", "text": "Every page works at phone width before it ships.", "why": "a page that only works on a laptop is a bug nobody reports." },
    { "key": "W4", "text": "Secrets never appear in code, logs or commits.", "why": "a key pushed once stays public, even after you delete it." },
    { "key": "W5", "text": "A new dependency comes with a one-line reason.", "why": "every package you add is code you now maintain." },
    { "key": "W6", "text": "An error a person sees says what to do next.", "why": "\"something went wrong\" is a support ticket waiting to happen." },
    { "key": "W7", "text": "Tests bind a local port, so a sandboxed agent asks the person instead of rewriting the tests.", "why": "a test rewritten into a fake passes inside the sandbox and proves nothing about the real server.", "tier": "must", "gate": "review" }
  ],
  "questions": [
    { "key": "Q1", "ask": "The app is built with", "choices": ["Next.js", "SvelteKit", "Rails", "Something else"], "detect": "framework" },
    { "key": "Q2", "ask": "It deploys to", "choices": ["Vercel", "Netlify", "Fly.io", "My own server"], "default": "Vercel" },
    { "key": "Q3", "ask": "People sign in with", "choices": ["Email links", "Passwords", "Google", "No sign-in yet"], "default": "Email links" },
    { "key": "Q4", "ask": "A change is done when", "choices": ["It runs on a preview link", "Its tests pass", "I have looked at it"], "default": "It runs on a preview link" }
  ],
  "items": [
    { "key": "1", "title": "Set up the test runner with one passing test", "cites": ["W1", "W2"], "criterion": "npm test runs the project's test runner and passes, with one real test in test/smoke.test.js that checks the app starts. Before this item npm test has no runner to start, so it fails.", "brief": "Files:\n- package.json (its test script)\n- test/smoke.test.js\nTest:\n- test/smoke.test.js checks the app starts\nUse the test runner your framework recommends, and give it a one-line reason (W5).", "check": "npm test" },
    { "key": "2", "title": "The home page works at phone width", "cites": ["W3"], "criterion": "The home page shows all of its content at 375 pixels wide without sideways scrolling. test/home-phone-width.test.js checks the page at that width and fails before the change.", "brief": "Files:\n- the home page\n- test/home-phone-width.test.js\nTest:\n- test/home-phone-width.test.js renders the home page 375 pixels wide and fails when anything scrolls sideways", "check": "npm test -- test/home-phone-width.test.js" },
    { "key": "3", "title": "Every change gets a preview link", "cites": ["Q2", "Q4"], "after": ["1"], "criterion": "Every pull request gets a preview link from the host the deploy question names. test/preview-config.test.js reads the deploy settings, checks that previews are turned on, and fails before the change.", "brief": "Files:\n- the host's deploy settings file\n- test/preview-config.test.js\nTest:\n- test/preview-config.test.js reads the deploy settings and fails when previews are off", "check": "npm test -- test/preview-config.test.js" }
  ]
}