vibeeval/property-based-testing
Property-based testing (PBT) patterns with fast-check (JS/TS), Hypothesis (Python), and gopter (Go). Generate random inputs, define invariants, shrink failures to minimal cases. Adapted from Trail of Bits. Use when testing pure functions, parsers, serializers, state machines, or any code where example-based tests miss edge cases.
npx skills add https://github.com/vibeeval/vibecosystem --skill property-based-testing
Instead of testing specific examples, define properties that must hold for ALL inputs. The framework generates hundreds of random inputs and finds the smallest failing case.
| Use Case | Property |
|----------|----------|
| Serialization roundtrip | deserialize(serialize(x)) === x |
| Sort function | Output is ordered AND contains same elements |
| Parser | Never crashes on any input |
| Encoder/decoder | decode(encode(x)) === x |
| State machine | Invariants hold after any sequence of operations |
| Math/financial | Associativity, commutativity, identity, bounds |
| API handler | Never returns 500 on valid input |
| Data transformation | Output schema matches specification |
npm install --save-dev fast-check
import fc from 'fast-check'
// Property: sorting is idempotent
test('sort is idempotent', () => {
fc.assert(
fc.property(fc.array(fc.integer()), (arr) => {
const sorted = [...arr].sort((a, b) => a - b)
const sortedTwice = [...sorted].sort((a, b) => a - b)
expect(sorted).toEqual(sortedTwice)
})
)
})
// Property: serialization roundtrip
test('JSON roundtrip preserves data', () => {
fc.assert(
fc.property(fc.jsonValue(), (value) => {
expect(JSON.parse(JSON.stringify(value))).toEqual(value)
})
)
})
// Generate valid email addresses
const emailArb = fc.tuple(
fc.stringOf(fc.constantFrom(...'abcdefghijklmnopqrstuvwxyz0123456789'.split('')), { minLength: 1 }),
fc.constantFrom('gmail.com', 'example.com', 'test.org')
).map(([local, domain]) => `${local}@${domain}`)
// Generate valid user objects
const userArb = fc.record({
id: fc.uuid(),
name: fc.string({ minLength: 1, maxLength: 100 }),
email: emailArb,
age: fc.integer({ min: 0, max: 150 }),
role: fc.constantFrom('admin', 'user', 'viewer')
})
// Generate valid but adversarial strings
const adversarialStringArb = fc.oneof(
fc.constant(''),
fc.constant(' '),
fc.constant('\0'),
fc.constant('<script>alert(1)</script>'),
fc.constant("Robert'); DROP TABLE users;--"),
fc.constant('../../../etc/passwd'),
fc.unicodeString(),
fc.string({ minLength: 10000, maxLength: 100000 }) // Very long
)
// Test a cache against a simple Map model
class CacheModel {
private model = new Map<string, string>()
set(key: string, value: string): void { this.model.set(key, value) }
get(key: string): string | undefined { return this.model.get(key) }
delete(key: string): void { this.model.delete(key) }
size(): number { return this.model.size }
}
const cacheCommands = [
fc.tuple(fc.string(), fc.string()).map(([k, v]) => ({
check: (model: CacheModel) => true,
run: (model: CacheModel, real: Cache) => {
model.set(k, v)
real.set(k, v)
expect(real.get(k)).toBe(model.get(k))
},
toString: () => `set(${k}, ${v})`
})),
fc.string().map((k) => ({
check: (model: CacheModel) => true,
run: (model: CacheModel, real: Cache) => {
model.delete(k)
real.delete(k)
expect(real.get(k)).toBe(model.get(k))
},
toString: () => `delete(${k})`
}))
]
test('cache behaves like Map', () => {
fc.assert(
fc.property(fc.commands(cacheCommands), (cmds) => {
const model = new CacheModel()
const real = new Cache()
fc.modelRun(() => ({ model, real }), cmds)
})
)
})
pip install hypothesis
from hypothesis import given, strategies as st, settings
@given(st.lists(st.integers()))
def test_sort_preserves_length(xs):
assert len(sorted(xs)) == len(xs)
@given(st.lists(st.integers()))
def test_sort_preserves_elements(xs):
assert sorted(sorted(xs)) == sorted(xs)
@given(st.text())
def test_encode_decode_roundtrip(s):
assert s.encode('utf-8').decode('utf-8') == s
# With settings
@settings(max_examples=1000, deadline=None)
@given(st.dictionaries(st.text(), st.integers()))
def test_dict_operations(d):
import json
assert json.loads(json.dumps(d)) == d
from hypothesis import strategies as st
# Valid email strategy
emails = st.builds(
lambda local, domain: f"{local}@{domain}",
local=st.from_regex(r'[a-z0-9]{1,20}', fullmatch=True),
domain=st.sampled_from(['gmail.com', 'example.com'])
)
# Valid user strategy
users = st.fixed_dictionaries({
'name': st.text(min_size=1, max_size=100),
'email': emails,
'age': st.integers(min_value=0, max_value=150),
'role': st.sampled_from(['admin', 'user', 'viewer'])
})
go get github.com/leanovate/gopter
func TestSortIdempotent(t *testing.T) {
properties := gopter.NewProperties(gopter.DefaultTestParameters())
properties.Property("sort is idempotent", prop.ForAll(
func(xs []int) bool {
sorted := make([]int, len(xs))
copy(sorted, xs)
sort.Ints(sorted)
sortedTwice := make([]int, len(sorted))
copy(sortedTwice, sorted)
sort.Ints(sortedTwice)
return reflect.DeepEqual(sorted, sortedTwice)
},
gen.SliceOf(gen.Int()),
))
properties.TestingRun(t)
}
| Property | Definition | Example |
|----------|-----------|---------|
| Identity | f(x, identity) === x | add(x, 0) === x |
| Commutativity | f(a, b) === f(b, a) | add(a, b) === add(b, a) |
| Associativity | f(f(a, b), c) === f(a, f(b, c)) | add(add(a, b), c) === add(a, add(b, c)) |
| Idempotency | f(f(x)) === f(x) | sort(sort(xs)) === sort(xs) |
| Roundtrip | g(f(x)) === x | decode(encode(x)) === x |
| Invariant | property(f(x)) === true | length(sort(xs)) === length(xs) |
| Property | Check |
|----------|-------|
| No crash | Function never throws for any valid input |
| Bounded output | Output size is proportional to input size |
| No mutation | Input is not modified by the function |
| Deterministic | Same input always produces same output |
| Monotonic | If a <= b then f(a) <= f(b) |
When a property fails, the framework automatically shrinks the failing input to the smallest case that still fails:
Original failing input: [482, -1, 0, 99, -384, 7, 42, 0, -1]
Shrunk to: [1, 0]
This tells you the bug is about: handling zero in a list with other elements
Tips:
{ endOnFailure: true }Inspired by Trail of Bits property-based-testing plugin.
Take vibeeval/property-based-testing from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.
The instructions reference pip, npm.
Without those the skill loads but fails at the first command.