e2elisted

# Phosphor e2e suite ```bash npm run test:e2e # full: build + all specs npm run build # rebuild after any main/renderer change npx playwright test # specs only (uses the existing build!) npx playwright test -g "sidebar" # one test by title substring npx playwright test --headed # watch it run ``` **Playwright runs `out/` — a stale build is the #1 source of confusing failures.** If a spec fails right after you changed code, rebuild first. ## How it works - `e2e/smoke.spec.ts` — 8 serial specs, 1 worker; each test launches its own Electron instance via `_electron.launch`. - `e2e/fixtures/pi-stub.cjs` — a ~350-line deterministic RPC "pi": scripted streamed reply, an edit tool call with a diff, an artifact tool call. No network, no API key. - Env contract (all gated on `!app.isPackaged`): - `PHOSPHOR_PI_STUB` — path to the stub; main spawns it instead of real pi - `PHOSPHOR_E2E_WORKSPACE` — skips the native (undriveable) folder picker - `PHOSPHOR_TEST_USER_DATA` — isolates electron-store prefs per test - `PI_CODING_AGENT_DIR` — pins pi's session dir to a scratch dir - The stub writes a real session JSONL into the **mangled session dir** for the workspace (pi's `--<cwd with / → ->--` layout) — that's what makes sessions discoverable by the sidebar scanner. The mangling is duplicated from `electron/pi/pi-paths.ts` in the stub; if a sidebar/scan test fails mysteriously, check the t