pj
0a71e05588
integers, strings, emails and edgeCaseText read the same rng from() does, so under the model policy an authored InputText typed the same value on every step while the seeded arm varied it. That is a silently different experiment, not just a silently different action space. Single-valued spans are exempt, because both policies then get the same value: between(7,7), a zero-length string, and a one-entry corpus. length(4,4) is still refused, since the length is pinned but each character is drawn from 62. Claude-Session: https://claude.ai/code/session_01A5KmftdEJ49A9z5mF5ESrX
@sanderling/spec
TypeScript spec API for sanderling, a property-based UI fuzzer for mobile and web apps.
Spec authors write properties (what the app must always or eventually do), extractors (structured state from the UI), and action generators (what sanderling is allowed to do). The sanderling CLI evaluates the spec in a loop against a running app.
Install
npm install --save-dev @sanderling/spec
Usage
import { extract, always, eventually, actions, weighted, taps, swipes, InputText, Tap } from "@sanderling/spec";
const loggedIn = extract((s) => !!s.ax.find("id:home-tab-bar"));
const balance = extract<number>((s) => (s.snapshots.balance as number) ?? 0);
const emailField = extract((s) => s.ax.find("id:email-field"));
const submitButton = extract((s) => s.ax.find("id:sign-in-button"));
export const properties = {
balanceNeverNegative: always(() => balance.current >= 0),
loginSucceeds: eventually(() => loggedIn.current).within(30, "seconds"),
};
const doLogin = actions(() => {
if (loggedIn.current) return [];
const email = emailField.current;
const submit = submitButton.current;
if (!email || !submit) return [];
return [InputText({ into: email, text: "[email protected]" }), Tap({ on: submit })];
});
export const actionsRoot = weighted(
[50, doLogin],
[10, taps],
[2, swipes],
);
Setup actions
Some action generators are not fuzz targets but preconditions: they drive the
app from a fresh state into the surface you actually want to fuzz (login,
onboarding, permission grants, seed data). Export them as setup instead of
mixing them into actionsRoot. The runner tries setup first; if it yields
no action, it falls through to actionsRoot. State regressing back across the
precondition (e.g. logout under fuzz) automatically re-engages setup.
const login = actions(() => {
if (loggedIn.current) return [];
return [InputText({ into: emailField.current!, text: "[email protected]" }), Tap({ on: submitButton.current! })];
});
export const setup = login;
export const actionsRoot = weighted([60, browse], [40, edit]);
(globalThis as { setup?: unknown }).setup = setup;
Setup is just an ActionGenerator; compose with actions, weighted, or
whenRoute exactly like the main pool.
Works identically across Android, iOS, and web targets.