Make a fresh clone work without submodules or a build step

A fresh `git clone && bun install` previously could not run `bun dev`
or `bun run test:e2e`:

- @executor-js/vite-plugin and @executor-js/react/console-routes only
  ship built `dist/` entrypoints, which used to require knowing to run
  `turbo run build` after installing. The root `prepare` hook now
  bundles both on every install (tsup, well under a second), so vite
  configs resolve on a fresh clone. (Pointing `exports.default` at the
  TS source instead would only work on Node >= 23.6 type-stripping —
  vite externalizes these imports from vite.config.ts, so Node itself
  has to load them.)
- testkit still imported `vendor/mcporter/dist`, which requires
  initializing the submodule AND building it with pnpm. Consume the
  fork from npm as @executor-js/mcporter instead, like e2e/ already
  does; the submodule stays for developing the fork itself (see
  vendor/README.md).

(An earlier version of this branch also injected an in-memory WorkOS
Vault client under EXECUTOR_E2E_STUB; the wire-level WorkOS emulator
(#931) deleted the stub layer and serves the real /vault/v1/kv routes,
so that workaround is no longer needed.)

Verified by cloning the repo fresh into /tmp: install, format, lint,
typecheck, and the full e2e suite (cloud + selfhost) all pass with no
submodules initialized and no manual build step.

Author: RhysSullivan

.gitignore

@@ -60,10 +60,11 @@ apps/desktop/resources/
 apps/cloud/.dev-db/
 apps/cloud/.e2e-db/
 
-# e2e suite: generated run artifacts + throwaway target state
+# e2e suite: generated run artifacts + throwaway target state (the * also
+# covers ad-hoc variants like .e2e-stub-db-manual from debugging boots)
 e2e/runs/
-apps/cloud/.e2e-stub-db/
-apps/host-selfhost/.e2e-data/
+apps/cloud/.e2e-stub-db*/
+apps/host-selfhost/.e2e-data*/
 
 # playwright e2e artifacts
 test-results/

README.md

@@ -126,6 +126,17 @@ bun dev
 
 The dev server starts at `http://127.0.0.1:4788`.
 
+### Tests
+
+```bash
+bun run test       # unit + integration suites
+bun run test:e2e   # full-stack e2e: boots the cloud and self-host apps and drives them
+```
+
+The browser e2e scenarios need Playwright's Chromium once per machine:
+`bunx playwright install chromium`. The git submodules under `vendor/` are
+optional — see [vendor/README.md](vendor/README.md).
+
 ## Community
 
 Join the Discord: [https://discord.gg/eF29HBHwM6](https://discord.gg/eF29HBHwM6)

e2e/src/surfaces/mcp.ts

@@ -1,7 +1,7 @@
-// MCP surface: the vendored mcporter fork as a programmatic MCP client, with
-// headless OAuth via the target's consent strategy. Session methods are
-// Effects; mcporter itself is promise-native underneath. Assertions are
-// vitest's job.
+// MCP surface: our mcporter fork (@executor-js/mcporter on npm; develop it in
+// the vendor/mcporter submodule) as a programmatic MCP client, with headless
+// OAuth via the target's consent strategy. Session methods are Effects;
+// mcporter itself is promise-native underneath. Assertions are vitest's job.
 import { createHash, randomBytes, randomUUID } from "node:crypto";
 import { mkdtempSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";

package.json

@@ -63,7 +63,7 @@
     "release:publish:packages:prepare": "bun run scripts/publish-packages.ts --prepare-only",
     "release:smoke:packages": "bun run scripts/smoke-test-packed.ts",
     "clean": "bun run scripts/clean.ts",
-    "prepare": "effect-language-service patch && effect-tsgo patch"
+    "prepare": "effect-language-service patch && effect-tsgo patch && bun run --cwd packages/core/vite-plugin build:bundle && bun run --cwd packages/react build"
   },
   "dependencies": {},
   "devDependencies": {

packages/core/vite-plugin/package.json

@@ -35,6 +35,7 @@
   },
   "scripts": {
     "build": "tsup && (tsc --declaration --emitDeclarationOnly --outDir dist --rootDir src || true)",
+    "build:bundle": "tsup",
     "typecheck": "tsgo --noEmit",
     "test": "vitest run",
     "typecheck:slow": "tsc --noEmit"

testkit/README.md

@@ -1,7 +1,8 @@
 # testkit (MCP surface) — MVP
 
-Black-box tests through the **MCP surface**, on the vendored mcporter fork
-(`vendor/mcporter`, with its headless `cookieConsentStrategy`). Every run emits a
+Black-box tests through the **MCP surface**, on our mcporter fork
+(`@executor-js/mcporter` on npm, with its headless `cookieConsentStrategy`;
+develop the fork itself in the `vendor/mcporter` submodule). Every run emits a
 **chat-transcript recording** (`run.json`) — the MCP surface *is* an agent-in-a-chat,
 so that's the natural shape.