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.
JavaScript agentic-codingai-agentsai-governanceautonomous-agentsclaude-codeclicode-reviewcodexcoding-agentsdeveloper-toolsgit-hooksllmlocal-firstmulti-agentsoftware-factoryspec-driven-developmentverificationvibe-coding
{
"blueprint": 1,
"name": "test-reliability",
"title": "Test reliability",
"version": 1,
"kind": "practice",
"by": "Pullboard",
"summary": "Rules we learned fixing flaky tests on our own board, ready for yours.",
"prefix": "TR",
"rules": [
{ "key": "TR1", "text": "No test waits on the wall clock or a fixed sleep.", "why": "a test that sleeps is slow on an idle machine and flaky on a busy one." },
{ "key": "TR2", "text": "A flake is pinned on a change only after a run without it.", "why": "a test that fails under load looks like a broken change until it fails without that change too." },
{ "key": "TR3", "text": "A flake fix is proven by repeated runs.", "why": "a flaky test already passes some of the time, so one green run after a fix proves nothing; only a run of passes shows the flake is gone.", "tier": "must", "gate": "the whole test suite, run as many times in a row as the run count question says, every run passing" }
],
"questions": [
{ "key": "Q1", "ask": "Tests run with", "choices": ["node:test", "Vitest", "Jest", "Mocha", "pytest", "Something else"], "detect": "test-runner" },
{ "key": "Q2", "ask": "A flake fix passes this many runs in a row", "choices": ["5", "10", "20"], "default": "10" }
],
"items": [
{ "key": "1", "title": "Find a flaky test and make it deterministic", "cites": ["TR1", "TR2"], "criterion": "A test that failed in repeated runs of the unchanged suite is found, named in the submit note, and fixed so the suite passes ten runs in a row with no sleep or wall-clock wait. Before the fix, ten repeated runs of npm test fail at least once.", "brief": "Files:\n- the flaky test and the code it waits on\nTest:\n- run npm test twenty times with nothing changed to find a test that fails some of the time (TR2)\n- replace its sleep or wall-clock wait with a fake clock or an awaited event (TR1), then run the check", "check": "set -e; for run in 1 2 3 4 5 6 7 8 9 10; do npm test; done" }
]
}