security: decrypt and sign dApp approvals in the popup (closes #157)

The dApp transaction and signature approval paths sent the user's plaintext
password to the background over runtime.sendMessage and decrypted there. Both
now decrypt in the popup, where the password is typed, and put only the signed
artifact on the wire: the raw signed transaction, or the signature. Neither the
password, the recovery phrase, the xprv nor the private key crosses the
messaging boundary any more. This matches what the popup-side eth_sendTransaction
path in confirmTx.js already did.

The popup runs the same sequence ethers' own sendTransaction() runs internally
(populateTransaction, then signTransaction), so nonce, gas, fee and chain id
population are unchanged. The background keeps broadcast and approval
resolution, and when the popup cannot produce an artifact it reports the error
over the same message so the requesting page still gets a failure rather than
hanging.

Moving the secret out of the background must not turn the background into a
blind relay, so it re-derives the signer from the artifact and checks it
against the approval it is holding before acting: shared/approvalVerify.js
asserts that a raw transaction is the approved transaction signed by the
approved address, and that a signature covers the approved payload and recovers
to the approved address.

A wrong password is now caught in the popup before anything is sent, so it
fails with an inline full-sentence error and leaves the pending approval alive
to retry; previously it reached the background and destroyed the approval.
Rejection still resolves with EIP-1193 code 4001, and approvals still survive
popup close and reopen.

Removes the four standing TODO(security) markers, now that the flaw is gone.

Makefile

@@ -1,4 +1,4 @@
-.PHONY: bootstrap setup install test lint fmt fmt-check check docker hooks build build-debug clean dev
+.PHONY: bootstrap setup install test test-e2e lint fmt fmt-check check docker hooks build build-debug clean dev
 
 # Standard targets are thin shims; the implementations live in script/
 # per the scripts-to-rule-them-all pattern (see the Entrypoints section
@@ -16,6 +16,10 @@ install:
 test:
 	@script/test
 
+# Browser end-to-end suite. Requires docker; not part of check.
+test-e2e:
+	@script/test-e2e
+
 lint:
 	@script/lint
 

README.md

@@ -73,6 +73,8 @@ provide:
   git pre-commit hook
 - `script/projectname` — print the project name (used for the Docker image tag)
 - `script/test` — run the test suite (jest)
+- `script/test-e2e` — run the browser end-to-end suite (docker required; see
+  [End-to-End Tests](#end-to-end-tests))
 - `script/lint` — run the linter
 - `script/fmt` — format all files (writes)
 - `script/fmt-check` — check formatting (read-only)
@@ -82,6 +84,66 @@ provide:
 - `script/precommit` — run by the git pre-commit hook; runs `script/check`
 - `script/install-precommit` — install the git pre-commit hook
 
+## End-to-End Tests
+
+`make test-e2e` builds `dist/chrome/` and drives the **real popup in a real
+Chrome**, loaded as an unpacked MV3 extension inside a pinned
+`mcr.microsoft.com/playwright` container (pinned by digest in `script/test-e2e`;
+docker is required and the suite fails loudly rather than skipping if it is
+unavailable). The suite lives in `tests/e2e/` and is driven by
+`playwright-core`, whose version must stay matched to the container's Playwright
+version — the browsers ship inside the image.
+
+It covers popup load, wallet creation through the UI, the Add Token screen and
+the transaction detail screen for an ERC-20 transfer. All outbound network is
+intercepted at the browser level and served from fixtures in
+`tests/e2e/network.js`, so the run is deterministic and fully offline;
+unrecognised outbound requests are reported as failures rather than silently
+allowed.
+
+That reporting has one bound worth knowing. Observation ends when the browser
+context is torn down, and nothing can watch traffic after that, so the run keeps
+collecting for a fixed grace period after the last test returns
+(`TRAILING_WATCH_MS` in `tests/e2e/run.js`, currently 1500ms) and then closes
+the context. A request whose _first_ dispatch falls after that window is never
+seen at all and cannot fail the run. In practice a request a test fires without
+awaiting reaches the route handler about 10ms later, and anything on a repeating
+timer gets observed on an earlier tick during the ~20s suite — but a one-shot
+call deliberately deferred past the window will escape.
+
+That interception covers the MV3 background service worker as well as the popup
+page, which it does not by default — `script/test-e2e` sets
+`PW_EXPERIMENTAL_SERVICE_WORKER_NETWORK_EVENTS=1` for it. Because that flag is
+experimental, the harness does not take it on trust. At launch it waits for the
+background worker's **own** startup request — the phishing blocklist fetch that
+`src/background/index.js` issues unconditionally — to arrive in the route
+handler, and aborts the entire suite if none does within 30 seconds
+(`tests/e2e/harness.js`). The check is passive on purpose: a synthetic probe
+fetched from inside the worker via `worker.evaluate()` was tried first and
+rejected, because evaluating in an extension service worker that early kills the
+worker outright, destroying the thing being measured. Observing traffic the
+extension already generates perturbs nothing. Losing the race fails closed — the
+suite refuses to run rather than passing quietly.
+
+As defence in depth, Chrome is also started with
+`--host-resolver-rules=MAP * ~NOTFOUND`, so a request that ever did slip past
+the route handler could not resolve a host at all. That only bounds the damage;
+detecting escaping traffic remains the canary's job. To see what is actually
+being intercepted, run with `E2E_TRACE_NETWORK=1` and every routed request is
+printed, tagged `[sw]` or `[page]`.
+
+**Any uncaught page error or `console.error` fails the run.** That is the point:
+a `ReferenceError` from a used-but-not-imported identifier is invisible to
+`make check` (`script/lint` is only `prettier --check`) but fatal in a browser,
+and this suite exists because exactly that class of bug shipped twice.
+
+`make test-e2e` is deliberately **not** part of `make check` or `make test`.
+`REPO_POLICIES.md` caps `make test` at 20 seconds and a browser suite does not
+fit; nothing in `tests/e2e/` is named `*.test.js`, so jest cannot pick it up
+either. It is also not wired into the Gitea workflow yet — docker-in-docker in
+CI is a separate question. Run it locally before changing anything under
+`src/popup/views/`.
+
 ## Rationale
 
 Common popular EVM wallets have become bloated with swap UIs, portfolio

TODO.md

@@ -15,7 +15,8 @@ other branch is in flight: the settings About well landed as #145 on 2026-07-26
 and scripts-to-rule-them-all landed as #148, so the `scripts/` directory
 question is resolved. Full policy file set present. `make check` verified
 passing on `main` at `23aeae4` on 2026-08-09. The 1.0.0 backlog is filed as
-#149-#168.
+#149-#168. A real-browser end-to-end suite (`make test-e2e`) now sits alongside
+`make check`, which cannot see a runtime `ReferenceError` in a popup view.
 
 # Next Step
 
@@ -27,8 +28,21 @@ review.
 
 # Completed Steps
 
+- 2026-08-09: dApp approval signing moved into the popup — the password no
+  longer crosses the extension messaging boundary; the background broadcasts and
+  resolves approvals only, and verifies the signed artifact against the approval
+  it holds (#157).
+- 2026-08-09: Containerized Chrome end-to-end harness (`make test-e2e` /
+  `script/test-e2e`) driving the real popup with all network intercepted, plus
+  the two used-but-not-imported crashes it caught: AddToken unreachable (#150)
+  and TransactionDetail broken for every ERC-20 transfer (#151). Harness
+  demonstrated failing before the fixes and passing after (#181). Interception
+  covers the MV3 background service worker, not just the popup page, and a
+  launch-time canary aborts the suite if worker traffic starts escaping.
 - 2026-08-09: Reviewed the repo end to end and filed the 1.0.0 backlog
   (#149-#168).
+- 2026-08-09: Test coverage for the address-poisoning defense in
+  `src/shared/transactions.js` (#160)
 - 2026-07-26: About well in settings with build info, repo link and the version
   click easter egg (#145); proper view navigation stack (#146).
 - 2026-07-07 Adopted scripts-to-rule-them-all: `script/` entrypoints, Makefile
@@ -52,10 +66,17 @@ review.
 
 # Future Steps
 
-- Fix the two `ReferenceError` crashes that make whole screens unreachable:
-  AddToken (#150) and TransactionDetail for every ERC-20 transfer (#151).
 - Add ESLint to `script/lint` (#152). `make check` is `prettier --check` only
   and cannot catch undefined identifiers, which is how #150 and #151 shipped.
+- Decide the libsodium backend that actually ships (#182) and delete the single
+  allowlist entry it owns in `tests/e2e/harness.js`.
+- Extend the end-to-end suite to the dApp approval signing path (EIP-1193
+  through the real content script, background worker and approval popup). That
+  path needs a CDP-based route to background-worker console output first:
+  Playwright exposes no error event for service workers, so an uncaught
+  exception in the worker cannot fail the run today (worker network traffic is
+  already covered). Decide separately whether docker-in-docker makes
+  `make test-e2e` runnable in the Gitea workflow.
 - Make the Firefox target functional: Chrome callback APIs are used against the
   promise-only `browser` namespace (#153).
 - Send and transaction-flow correctness: gas fee excluded from the

package.json

@@ -16,6 +16,7 @@
         "@tailwindcss/cli": "^4.2.1",
         "esbuild": "^0.27.3",
         "jest": "^30.2.0",
+        "playwright-core": "1.56.0",
         "prettier": "^3.8.1",
         "tailwindcss": "^4.2.1"
     },

script/test-e2e

@@ -0,0 +1,64 @@
+#!/bin/sh
+# script/test-e2e: build the extension and drive the real popup in a real
+# Chromium inside a pinned container. Our own extension to
+# scripts-to-rule-them-all.
+#
+# Deliberately NOT called by script/check or script/test: REPO_POLICIES.md
+# caps make test at 20 seconds and a browser suite does not fit. Run it
+# yourself before touching popup views; it is the only check that can see
+# a used-but-not-imported identifier blow up at runtime.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+# mcr.microsoft.com/playwright:v1.56.0-noble, 2026-08-09
+#
+# The playwright-core devDependency is pinned to the matching Playwright
+# version (1.56.0) and the two must be bumped together: the browsers ship
+# inside this image, and playwright-core looks for the exact browser
+# revision its own version expects. A mismatch fails at launch.
+IMAGE="mcr.microsoft.com/playwright@sha256:35246d87a7c88ea9b771c65d33171b2611b02a8253b4b12ce6f94376c55f99f2"
+
+main() {
+    cd "$ROOT"
+
+    if ! command -v docker >/dev/null 2>&1; then
+        echo "test-e2e: docker is required to run the e2e suite" >&2
+        exit 1
+    fi
+
+    echo "Building extension for e2e..."
+    yarn run build 2>&1
+
+    echo "Running e2e suite in the pinned Playwright container..."
+    # --ipc=host: Chromium's shared-memory needs more than the default
+    #   64MB /dev/shm or renderers crash.
+    # --user: keep files the suite touches owned by the caller, not root.
+    # HOME=/tmp: the mapped uid has no home directory in the image.
+    # PW_EXPERIMENTAL_SERVICE_WORKER_NETWORK_EVENTS=1: without it,
+    #   ctx.route() intercepts page requests only, and every fetch made by
+    #   the MV3 background service worker — including the phishing
+    #   blocklist fetch that src/background/index.js issues at worker
+    #   startup — goes to the real internet. The flag is experimental and
+    #   Playwright may drop or rename it. It cannot break silently: the
+    #   harness probes service-worker interception at launch and aborts
+    #   the whole suite if it is not in effect (see the interception
+    #   canary in tests/e2e/harness.js). If a future Playwright removes
+    #   the flag, that probe is what will fail, and the fix is either a
+    #   replacement mechanism or an honest downgrade of the isolation
+    #   claim in tests/e2e/network.js and README.md — not deleting the
+    #   probe. The image is pinned by digest, so this can only ever bite
+    #   on a deliberate bump.
+    docker run --rm \
+        --ipc=host \
+        --user "$(id -u):$(id -g)" \
+        -e HOME=/tmp \
+        -e PW_EXPERIMENTAL_SERVICE_WORKER_NETWORK_EVENTS=1 \
+        -e "E2E_TRACE_NETWORK=${E2E_TRACE_NETWORK:-0}" \
+        -v "$ROOT:/work" \
+        -w /work \
+        "$IMAGE" \
+        node tests/e2e/run.js
+}
+
+main "$@"

src/background/index.js

@@ -5,7 +5,6 @@
 const { DEFAULT_RPC_URL } = require("../shared/constants");
 const { SUPPORTED_CHAIN_IDS, networkByChainId } = require("../shared/networks");
 const { onChainSwitch } = require("../shared/chainSwitch");
-const { getBytes } = require("ethers");
 const {
     state,
     loadState,
@@ -14,8 +13,7 @@ const {
 } = require("../shared/state");
 const { refreshBalances, getProvider } = require("../shared/balances");
 const { debugFetch } = require("../shared/log");
-const { decryptWithPassword } = require("../shared/vault");
-const { getSignerForAddress } = require("../shared/wallet");
+const { verifySignedTx, verifySignature } = require("../shared/approvalVerify");
 const {
     isPhishingDomain,
     updatePhishingList,
@@ -725,39 +723,28 @@ runtime.onMessage.addListener((msg, sender, sendResponse) => {
             return true;
         }
 
+        // The popup signs; it reports back here when it could not. Fail the
+        // request the same way this handler used to when it did the signing.
+        if (msg.error) {
+            approval.resolve({ error: { message: msg.error } });
+            sendResponse({ error: msg.error });
+            return false;
+        }
+
         (async () => {
             try {
                 await loadState();
                 const activeAddress = await getActiveAddress();
-                let wallet, addrIndex;
-                for (const w of state.wallets) {
-                    for (let i = 0; i < w.addresses.length; i++) {
-                        if (w.addresses[i].address === activeAddress) {
-                            wallet = w;
-                            addrIndex = i;
-                            break;
-                        }
-                    }
-                    if (wallet) break;
-                }
-                if (!wallet) throw new Error("Wallet not found");
-                // TODO(security): Move decryption to popup to avoid sending password via runtime.sendMessage
-                let decrypted = await decryptWithPassword(
-                    wallet.encryptedSecret,
-                    msg.password,
+                // The popup holds the secret, but the background stays the
+                // authority on what is broadcast: the raw transaction must be
+                // the approved one, signed by the approved address.
+                verifySignedTx(
+                    msg.rawSignedTx,
+                    approval.txParams,
+                    activeAddress,
                 );
-                const signer = getSignerForAddress(
-                    wallet,
-                    addrIndex,
-                    decrypted,
-                );
-                // Best-effort: clear decrypted secret after use.
-                // Note: JS strings are immutable; this nulls the reference but
-                // the original string may persist in memory until GC.
-                decrypted = null;
                 const provider = getProvider(state.rpcUrl);
-                const connected = signer.connect(provider);
-                const tx = await connected.sendTransaction(approval.txParams);
+                const tx = await provider.broadcastTransaction(msg.rawSignedTx);
                 approval.resolve({ txHash: tx.hash });
                 sendResponse({ txHash: tx.hash });
             } catch (e) {
@@ -784,55 +771,23 @@ runtime.onMessage.addListener((msg, sender, sendResponse) => {
             return true;
         }
 
+        // The popup signs; it reports back here when it could not. Fail the
+        // request the same way this handler used to when it did the signing.
+        if (msg.error) {
+            approval.resolve({ error: { message: msg.error } });
+            sendResponse({ error: msg.error });
+            return false;
+        }
+
         (async () => {
             try {
-                await loadState();
                 const activeAddress = await getActiveAddress();
-                let wallet, addrIndex;
-                for (const w of state.wallets) {
-                    for (let i = 0; i < w.addresses.length; i++) {
-                        if (w.addresses[i].address === activeAddress) {
-                            wallet = w;
-                            addrIndex = i;
-                            break;
-                        }
-                    }
-                    if (wallet) break;
-                }
-                if (!wallet) throw new Error("Wallet not found");
-                // TODO(security): Move decryption to popup to avoid sending password via runtime.sendMessage
-                let decrypted = await decryptWithPassword(
-                    wallet.encryptedSecret,
-                    msg.password,
-                );
-                const signer = getSignerForAddress(
-                    wallet,
-                    addrIndex,
-                    decrypted,
-                );
-                // Best-effort: clear decrypted secret after use.
-                // Note: JS strings are immutable; this nulls the reference but
-                // the original string may persist in memory until GC.
-                decrypted = null;
-
-                const sp = approval.signParams;
-                let signature;
-
-                if (sp.method === "personal_sign" || sp.method === "eth_sign") {
-                    signature = await signer.signMessage(getBytes(sp.message));
-                } else {
-                    // eth_signTypedData_v4 / eth_signTypedData
-                    const typedData = JSON.parse(sp.typedData);
-                    const { domain, types, message } = typedData;
-                    // ethers handles EIP712Domain internally
-                    delete types.EIP712Domain;
-                    signature = await signer.signTypedData(
-                        domain,
-                        types,
-                        message,
-                    );
-                }
-
+                // The popup holds the secret, but the background stays the
+                // authority on what is handed back to the page: the signature
+                // must cover the approved payload and recover to the approved
+                // address.
+                const signature = msg.signature;
+                verifySignature(approval.signParams, signature, activeAddress);
                 approval.resolve({ signature });
                 sendResponse({ signature });
             } catch (e) {

src/popup/views/addToken.js

@@ -1,4 +1,4 @@
-const { $, showFlash, goBack } = require("./helpers");
+const { $, showView, showFlash, goBack } = require("./helpers");
 const { getTopTokens } = require("../../shared/tokenList");
 const { state, saveState } = require("../../shared/state");
 const { lookupTokenInfo } = require("../../shared/balances");

src/popup/views/approval.js

@@ -9,10 +9,19 @@ const {
     attachCopyHandlers,
 } = require("./helpers");
 const { state, saveState, currentNetwork } = require("../../shared/state");
-const { formatEther, formatUnits, Interface, toUtf8String } = require("ethers");
+const {
+    formatEther,
+    formatUnits,
+    getBytes,
+    Interface,
+    toUtf8String,
+} = require("ethers");
 const { getPrice, formatUsd } = require("../../shared/prices");
 const { ERC20_ABI } = require("../../shared/constants");
 const { TOKEN_BY_ADDRESS } = require("../../shared/tokenList");
+const { decryptWithPassword } = require("../../shared/vault");
+const { getSignerForAddress } = require("../../shared/wallet");
+const { getProvider } = require("../../shared/balances");
 const txStatus = require("./txStatus");
 const uniswap = require("../../shared/uniswap");
 const runtime =
@@ -153,6 +162,8 @@ function showTxApproval(details) {
         details.isPhishingDomain,
     );
 
+    pendingTxParams = details.txParams;
+
     const toAddr = details.txParams.to;
     const token = toAddr ? TOKEN_BY_ADDRESS.get(toAddr.toLowerCase()) : null;
     const ethValue = formatEther(details.txParams.value || "0");
@@ -326,6 +337,7 @@ function showSignApproval(details) {
     );
 
     const sp = details.signParams;
+    pendingSignParams = sp;
 
     $("approve-sign-hostname").textContent = details.hostname;
     $("approve-sign-from").innerHTML = approvalAddressHtml(sp.from);
@@ -401,6 +413,36 @@ function show(id) {
 
 let approvalId = null;
 let pendingTxDetails = null;
+// The exact parameters shown to the user, kept so the popup signs what it
+// displayed rather than re-fetching anything at approval time. Both are
+// repopulated by show() when the popup is closed and reopened.
+let pendingTxParams = null;
+let pendingSignParams = null;
+
+// Approve buttons stay disabled and muted while the popup derives the key and
+// signs, which is slow enough (Argon2id) that a double click is likely.
+function setTxButtonBusy(busy) {
+    $("btn-approve-tx").disabled = busy;
+    $("btn-approve-tx").classList.toggle("text-muted", busy);
+}
+
+function setSignButtonBusy(busy) {
+    $("btn-approve-sign").disabled = busy;
+    $("btn-approve-sign").classList.toggle("text-muted", busy);
+}
+
+// Locate the wallet and the address index owning the currently active
+// address. Returns null when no wallet holds it.
+function findActiveWallet() {
+    for (const wallet of state.wallets) {
+        for (let i = 0; i < wallet.addresses.length; i++) {
+            if (wallet.addresses[i].address === state.activeAddress) {
+                return { wallet, addrIndex: i };
+            }
+        }
+    }
+    return null;
+}
 
 function init(ctx) {
     $("approve-remember").addEventListener("change", async () => {
@@ -430,25 +472,78 @@ function init(ctx) {
         window.close();
     });
 
-    $("btn-approve-tx").addEventListener("click", () => {
-        const password = $("approve-tx-password").value;
+    $("btn-approve-tx").addEventListener("click", async () => {
+        let password = $("approve-tx-password").value;
         if (!password) {
             showError("approve-tx-error", "Please enter your password.");
             return;
         }
         hideError("approve-tx-error");
-        $("btn-approve-tx").disabled = true;
-        $("btn-approve-tx").classList.add("text-muted");
+        setTxButtonBusy(true);
 
-        runtime.sendMessage(
-            {
+        const active = findActiveWallet();
+        if (!active) {
+            password = null;
+            showError(
+                "approve-tx-error",
+                "No wallet was found for the active address.",
+            );
+            setTxButtonBusy(false);
+            return;
+        }
+
+        // Decrypt here, in the popup. The password must never cross the
+        // extension messaging boundary; only the signed transaction does.
+        let decryptedSecret;
+        try {
+            decryptedSecret = await decryptWithPassword(
+                active.wallet.encryptedSecret,
+                password,
+            );
+        } catch {
+            showError(
+                "approve-tx-error",
+                "That password is incorrect. Please try again.",
+            );
+            setTxButtonBusy(false);
+            return;
+        } finally {
+            // Best-effort: drop the password as soon as the key derivation
+            // is done. Note that JS strings are immutable; this clears the
+            // reference but the original string may persist until GC.
+            password = null;
+        }
+
+        const payload = {
             type: "AUTISTMASK_TX_RESPONSE",
             id: approvalId,
             approved: true,
-                // TODO(security): Move decryption to popup to avoid sending password via runtime.sendMessage
-                password: password,
-            },
-            (response) => {
+        };
+        try {
+            const signer = getSignerForAddress(
+                active.wallet,
+                active.addrIndex,
+                decryptedSecret,
+            );
+            const provider = getProvider(state.rpcUrl);
+            const connected = signer.connect(provider);
+            // This is the sequence ethers' own sendTransaction() runs
+            // internally, so nonce, gas, fee and chain id population are
+            // identical to when the background did the signing.
+            const populated =
+                await connected.populateTransaction(pendingTxParams);
+            delete populated.from;
+            payload.rawSignedTx = await connected.signTransaction(populated);
+        } catch (e) {
+            payload.error =
+                e.shortMessage || e.message || "Transaction signing failed.";
+        } finally {
+            // Best-effort: clear the decrypted secret after use, with the
+            // same immutability caveat as the password above.
+            decryptedSecret = null;
+        }
+
+        runtime.sendMessage(payload, (response) => {
             if (response && response.txHash) {
                 txStatus.showWait(pendingTxDetails, response.txHash);
             } else {
@@ -456,8 +551,7 @@ function init(ctx) {
                     (response && response.error) || "Transaction failed.";
                 txStatus.showError(pendingTxDetails, null, msg);
             }
-            },
-        );
+        });
     });
 
     $("btn-reject-tx").addEventListener("click", () => {
@@ -469,36 +563,93 @@ function init(ctx) {
         window.close();
     });
 
-    $("btn-approve-sign").addEventListener("click", () => {
-        const password = $("approve-sign-password").value;
+    $("btn-approve-sign").addEventListener("click", async () => {
+        let password = $("approve-sign-password").value;
         if (!password) {
             showError("approve-sign-error", "Please enter your password.");
             return;
         }
         hideError("approve-sign-error");
-        $("btn-approve-sign").disabled = true;
-        $("btn-approve-sign").classList.add("text-muted");
+        setSignButtonBusy(true);
 
-        runtime.sendMessage(
-            {
+        const active = findActiveWallet();
+        if (!active) {
+            password = null;
+            showError(
+                "approve-sign-error",
+                "No wallet was found for the active address.",
+            );
+            setSignButtonBusy(false);
+            return;
+        }
+
+        // Decrypt here, in the popup. The password must never cross the
+        // extension messaging boundary; only the signature does.
+        let decryptedSecret;
+        try {
+            decryptedSecret = await decryptWithPassword(
+                active.wallet.encryptedSecret,
+                password,
+            );
+        } catch {
+            showError(
+                "approve-sign-error",
+                "That password is incorrect. Please try again.",
+            );
+            setSignButtonBusy(false);
+            return;
+        } finally {
+            // Best-effort: drop the password as soon as the key derivation
+            // is done. Note that JS strings are immutable; this clears the
+            // reference but the original string may persist until GC.
+            password = null;
+        }
+
+        const payload = {
             type: "AUTISTMASK_SIGN_RESPONSE",
             id: approvalId,
             approved: true,
-                // TODO(security): Move decryption to popup to avoid sending password via runtime.sendMessage
-                password: password,
-            },
-            (response) => {
+        };
+        try {
+            const signer = getSignerForAddress(
+                active.wallet,
+                active.addrIndex,
+                decryptedSecret,
+            );
+            const sp = pendingSignParams;
+            if (sp.method === "personal_sign" || sp.method === "eth_sign") {
+                payload.signature = await signer.signMessage(
+                    getBytes(sp.message),
+                );
+            } else {
+                // eth_signTypedData_v4 / eth_signTypedData
+                const typedData = JSON.parse(sp.typedData);
+                const { domain, types, message } = typedData;
+                // ethers handles EIP712Domain internally
+                delete types.EIP712Domain;
+                payload.signature = await signer.signTypedData(
+                    domain,
+                    types,
+                    message,
+                );
+            }
+        } catch (e) {
+            payload.error = e.shortMessage || e.message || "Signing failed.";
+        } finally {
+            // Best-effort: clear the decrypted secret after use, with the
+            // same immutability caveat as the password above.
+            decryptedSecret = null;
+        }
+
+        runtime.sendMessage(payload, (response) => {
             if (response && response.signature) {
                 window.close();
             } else {
-                    const msg =
-                        (response && response.error) || "Signing failed.";
+                const msg = (response && response.error) || "Signing failed.";
                 showError("approve-sign-error", msg);
-                    $("btn-approve-sign").disabled = false;
-                    $("btn-approve-sign").classList.remove("text-muted");
+                setSignButtonBusy(false);
             }
-            },
-        );
+        });
     });
 
     $("btn-reject-sign").addEventListener("click", () => {

src/popup/views/transactionDetail.js

@@ -7,6 +7,7 @@ const {
     showFlash,
     flashCopyFeedback,
     addressTitle,
+    addressDotHtml,
     escapeHtml,
     isoDate,
     timeAgo,

src/shared/approvalVerify.js

@@ -0,0 +1,124 @@
+// Verification of the signed artifacts produced by the approval popup.
+//
+// Signing happens in the popup, where the password is entered; the background
+// only broadcasts the raw transaction and resolves the pending approval back
+// to the requesting page. So that moving the signing out of the background
+// does not turn the background into a blind relay, the background re-derives
+// the signer from the artifact and checks it against the approval it is
+// holding before acting on it. All recovery is delegated to ethers.
+//
+// Every failure message is a full sentence, because these strings are shown to
+// the user and returned to the dApp.
+
+const {
+    Transaction,
+    getAddress,
+    getBytes,
+    verifyMessage,
+    verifyTypedData,
+} = require("ethers");
+
+// Case-insensitive address comparison that tolerates absent values on either
+// side. Two absent addresses compare equal (contract creation has no `to`).
+function sameAddress(a, b) {
+    const aMissing = a === null || a === undefined || a === "";
+    const bMissing = b === null || b === undefined || b === "";
+    if (aMissing || bMissing) return aMissing && bMissing;
+    try {
+        return getAddress(a) === getAddress(b);
+    } catch {
+        return String(a).toLowerCase() === String(b).toLowerCase();
+    }
+}
+
+// Normalize a transaction value (hex string, decimal string, number or
+// bigint) to a bigint. An absent value is zero, matching ethers.
+function normalizeValue(v) {
+    if (v === null || v === undefined || v === "") return 0n;
+    return BigInt(v);
+}
+
+// Normalize call data to a lowercase hex string. Absent data is "0x".
+function normalizeData(v) {
+    if (v === null || v === undefined || v === "" || v === "0x") return "0x";
+    return String(v).toLowerCase();
+}
+
+// Assert that a raw signed transaction is the transaction the user approved,
+// signed by the address the approval was raised for. Returns the parsed
+// ethers Transaction on success, throws otherwise.
+function verifySignedTx(rawSignedTx, txParams, expectedFrom) {
+    if (typeof rawSignedTx !== "string" || !rawSignedTx.startsWith("0x")) {
+        throw new Error("The signed transaction is missing or malformed.");
+    }
+
+    let parsed;
+    try {
+        parsed = Transaction.from(rawSignedTx);
+    } catch {
+        throw new Error("The signed transaction could not be decoded.");
+    }
+
+    if (!parsed.from) {
+        throw new Error("The signed transaction carries no valid signature.");
+    }
+    if (!sameAddress(parsed.from, expectedFrom)) {
+        throw new Error(
+            "The signed transaction was signed by a different address than the one that was approved.",
+        );
+    }
+    if (!sameAddress(parsed.to, txParams.to)) {
+        throw new Error(
+            "The signed transaction does not go to the approved recipient.",
+        );
+    }
+    if (normalizeValue(parsed.value) !== normalizeValue(txParams.value)) {
+        throw new Error(
+            "The signed transaction does not carry the approved value.",
+        );
+    }
+    if (normalizeData(parsed.data) !== normalizeData(txParams.data)) {
+        throw new Error(
+            "The signed transaction does not carry the approved call data.",
+        );
+    }
+
+    return parsed;
+}
+
+// Assert that a signature over the approved message or typed data was
+// produced by the address the approval was raised for. Returns the recovered
+// address on success, throws otherwise.
+function verifySignature(signParams, signature, expectedFrom) {
+    if (typeof signature !== "string" || !signature.startsWith("0x")) {
+        throw new Error("The signature is missing or malformed.");
+    }
+
+    let recovered;
+    try {
+        if (
+            signParams.method === "personal_sign" ||
+            signParams.method === "eth_sign"
+        ) {
+            recovered = verifyMessage(getBytes(signParams.message), signature);
+        } else {
+            const typedData = JSON.parse(signParams.typedData);
+            const { domain, types, message } = typedData;
+            // ethers derives EIP712Domain itself and rejects it as an input.
+            delete types.EIP712Domain;
+            recovered = verifyTypedData(domain, types, message, signature);
+        }
+    } catch {
+        throw new Error("The signature could not be verified.");
+    }
+
+    if (!sameAddress(recovered, expectedFrom)) {
+        throw new Error(
+            "The signature was produced by a different address than the one that was approved.",
+        );
+    }
+
+    return recovered;
+}
+
+module.exports = { verifySignedTx, verifySignature, sameAddress };

tests/approvalVerify.test.js

@@ -0,0 +1,355 @@
+const { Network, Transaction, Wallet } = require("ethers");
+const {
+    verifySignedTx,
+    verifySignature,
+    sameAddress,
+} = require("../src/shared/approvalVerify");
+const { getSignerForAddress } = require("../src/shared/wallet");
+
+// Fixed test keys — never used for anything but these tests.
+const SIGNER_KEY =
+    "0x59c6995e998f97a5a0044966f0945389dc9e86dae88c7a8412f4603b6b78690d";
+const OTHER_KEY =
+    "0x5de4111afa1a4b94908f83103eb1f1706367c2e68ca870fc3fb9a804cdab365a";
+
+const signer = new Wallet(SIGNER_KEY);
+const other = new Wallet(OTHER_KEY);
+
+const RECIPIENT = "0x66133E8ea0f5D1d612D2502a968757D1048c214a";
+const OTHER_RECIPIENT = "0xdAC17F958D2ee523a2206206994597C13D831ec7";
+
+// Approved parameters as a dApp would supply them over eth_sendTransaction.
+const TX_PARAMS = {
+    from: signer.address,
+    to: RECIPIENT,
+    value: "0x2386f26fc10000",
+    data: "0xdeadbeef",
+    gas: "0x5208",
+};
+
+// Build a signable transaction from approved params. The popup does the same
+// thing through populateTransaction(); here the fields are fixed so the test
+// needs no provider.
+function txFor(params) {
+    return {
+        chainId: 1,
+        nonce: 7,
+        gasLimit: 100000n,
+        maxFeePerGas: 2000000000n,
+        maxPriorityFeePerGas: 1000000000n,
+        type: 2,
+        to: params.to,
+        value: params.value === undefined ? 0n : BigInt(params.value),
+        data: params.data || "0x",
+    };
+}
+
+async function signedFor(params, withWallet) {
+    return (withWallet || signer).signTransaction(txFor(params));
+}
+
+describe("sameAddress", () => {
+    test("compares checksummed and lowercase forms as equal", () => {
+        expect(sameAddress(RECIPIENT, RECIPIENT.toLowerCase())).toBe(true);
+    });
+
+    test("treats two absent addresses as equal (contract creation)", () => {
+        expect(sameAddress(null, undefined)).toBe(true);
+        expect(sameAddress("", null)).toBe(true);
+    });
+
+    test("treats one absent address as unequal", () => {
+        expect(sameAddress(RECIPIENT, null)).toBe(false);
+        expect(sameAddress(null, RECIPIENT)).toBe(false);
+    });
+
+    test("does not throw on values that are not addresses", () => {
+        expect(sameAddress("not-an-address", RECIPIENT)).toBe(false);
+    });
+});
+
+describe("verifySignedTx", () => {
+    test("accepts the approved transaction signed by the approved address", async () => {
+        const raw = await signedFor(TX_PARAMS);
+        const parsed = verifySignedTx(raw, TX_PARAMS, signer.address);
+        expect(parsed.from).toBe(signer.address);
+        expect(parsed.hash).toBe(Transaction.from(raw).hash);
+    });
+
+    test("accepts a contract creation with no recipient", async () => {
+        const params = { to: undefined, value: "0x0", data: "0x600160005500" };
+        const raw = await signedFor(params);
+        expect(() => verifySignedTx(raw, params, signer.address)).not.toThrow();
+    });
+
+    test("accepts an absent value as zero", async () => {
+        const approved = { to: RECIPIENT, data: "0x" };
+        const raw = await signedFor(approved);
+        expect(() =>
+            verifySignedTx(raw, approved, signer.address),
+        ).not.toThrow();
+    });
+
+    test("accepts call data whose case differs from the approval", async () => {
+        const approved = { to: RECIPIENT, value: "0x0", data: "0xDEADBEEF" };
+        const raw = await signedFor(approved);
+        expect(() =>
+            verifySignedTx(raw, approved, signer.address),
+        ).not.toThrow();
+    });
+
+    test("rejects a swapped recipient", async () => {
+        const raw = await signedFor({
+            ...TX_PARAMS,
+            to: OTHER_RECIPIENT,
+        });
+        expect(() => verifySignedTx(raw, TX_PARAMS, signer.address)).toThrow(
+            /approved recipient/,
+        );
+    });
+
+    test("rejects an inflated value", async () => {
+        const raw = await signedFor({
+            ...TX_PARAMS,
+            value: "0x4563918244f40000",
+        });
+        expect(() => verifySignedTx(raw, TX_PARAMS, signer.address)).toThrow(
+            /approved value/,
+        );
+    });
+
+    test("rejects substituted call data", async () => {
+        const raw = await signedFor({ ...TX_PARAMS, data: "0xc0ffee" });
+        expect(() => verifySignedTx(raw, TX_PARAMS, signer.address)).toThrow(
+            /approved call data/,
+        );
+    });
+
+    test("rejects a transaction signed by a different address", async () => {
+        const raw = await signedFor(TX_PARAMS, other);
+        expect(() => verifySignedTx(raw, TX_PARAMS, signer.address)).toThrow(
+            /different address/,
+        );
+    });
+
+    test("rejects an unsigned transaction", () => {
+        const unsigned = Transaction.from(txFor(TX_PARAMS)).unsignedSerialized;
+        expect(() =>
+            verifySignedTx(unsigned, TX_PARAMS, signer.address),
+        ).toThrow(/no valid signature/);
+    });
+
+    test("rejects a missing or malformed payload", () => {
+        expect(() =>
+            verifySignedTx(undefined, TX_PARAMS, signer.address),
+        ).toThrow(/missing or malformed/);
+        expect(() => verifySignedTx("nope", TX_PARAMS, signer.address)).toThrow(
+            /missing or malformed/,
+        );
+        expect(() =>
+            verifySignedTx("0xc0ffee", TX_PARAMS, signer.address),
+        ).toThrow(/could not be decoded/);
+    });
+
+    test("every rejection message is a full sentence", async () => {
+        const raw = await signedFor({ ...TX_PARAMS, to: OTHER_RECIPIENT });
+        try {
+            verifySignedTx(raw, TX_PARAMS, signer.address);
+            throw new Error("expected a rejection");
+        } catch (e) {
+            expect(e.message).toMatch(/^[A-Z].*\.$/);
+        }
+    });
+});
+
+const TYPED_DATA = JSON.stringify({
+    domain: {
+        name: "AutistMask Test",
+        version: "1",
+        chainId: 1,
+        verifyingContract: OTHER_RECIPIENT,
+    },
+    primaryType: "Mail",
+    types: {
+        EIP712Domain: [
+            { name: "name", type: "string" },
+            { name: "version", type: "string" },
+            { name: "chainId", type: "uint256" },
+            { name: "verifyingContract", type: "address" },
+        ],
+        Mail: [
+            { name: "from", type: "address" },
+            { name: "to", type: "address" },
+            { name: "contents", type: "string" },
+        ],
+    },
+    message: {
+        from: signer.address,
+        to: RECIPIENT,
+        contents: "hello",
+    },
+});
+
+describe("verifySignature", () => {
+    // "Hello AutistMask" as the hex string a dApp passes to personal_sign.
+    const MESSAGE = "0x48656c6c6f204175746973744d61736b";
+    const personalParams = {
+        method: "personal_sign",
+        message: MESSAGE,
+        from: signer.address,
+    };
+    const typedParams = {
+        method: "eth_signTypedData_v4",
+        typedData: TYPED_DATA,
+        from: signer.address,
+    };
+
+    async function signPersonal(withWallet) {
+        return (withWallet || signer).signMessage(
+            Buffer.from(MESSAGE.slice(2), "hex"),
+        );
+    }
+
+    async function signTyped(withWallet) {
+        const { domain, types, message } = JSON.parse(TYPED_DATA);
+        delete types.EIP712Domain;
+        return (withWallet || signer).signTypedData(domain, types, message);
+    }
+
+    test("accepts a personal_sign signature from the approved address", async () => {
+        const signature = await signPersonal();
+        expect(verifySignature(personalParams, signature, signer.address)).toBe(
+            signer.address,
+        );
+    });
+
+    test("accepts an eth_sign signature the same way", async () => {
+        const signature = await signPersonal();
+        const params = { ...personalParams, method: "eth_sign" };
+        expect(() =>
+            verifySignature(params, signature, signer.address),
+        ).not.toThrow();
+    });
+
+    test("accepts a typed data signature from the approved address", async () => {
+        const signature = await signTyped();
+        expect(verifySignature(typedParams, signature, signer.address)).toBe(
+            signer.address,
+        );
+    });
+
+    test("does not mutate the approved typed data while verifying", async () => {
+        const signature = await signTyped();
+        const before = typedParams.typedData;
+        verifySignature(typedParams, signature, signer.address);
+        expect(typedParams.typedData).toBe(before);
+        expect(
+            JSON.parse(typedParams.typedData).types.EIP712Domain,
+        ).toBeDefined();
+    });
+
+    test("rejects a personal_sign signature from a different address", async () => {
+        const signature = await signPersonal(other);
+        expect(() =>
+            verifySignature(personalParams, signature, signer.address),
+        ).toThrow(/different address/);
+    });
+
+    test("rejects a typed data signature from a different address", async () => {
+        const signature = await signTyped(other);
+        expect(() =>
+            verifySignature(typedParams, signature, signer.address),
+        ).toThrow(/different address/);
+    });
+
+    test("rejects a signature over a different message", async () => {
+        const signature = await signer.signMessage(
+            Buffer.from("00112233", "hex"),
+        );
+        expect(() =>
+            verifySignature(personalParams, signature, signer.address),
+        ).toThrow(/different address/);
+    });
+
+    test("rejects a missing or malformed signature", async () => {
+        expect(() =>
+            verifySignature(personalParams, undefined, signer.address),
+        ).toThrow(/missing or malformed/);
+        expect(() =>
+            verifySignature(personalParams, "0x1234", signer.address),
+        ).toThrow(/could not be verified/);
+    });
+});
+
+// End-to-end over the messaging boundary, without a browser: run the exact
+// sequence the approval popup runs, then hand the artifact to the exact check
+// the background runs before it broadcasts or resolves. Only what the popup
+// puts on the wire is passed along, so this also pins down that the wire
+// payload is sufficient on its own.
+describe("popup signing sequence to background verification", () => {
+    // Stand-in for the JSON-RPC provider. populateTransaction only needs the
+    // nonce, the gas estimate, the network and the fee data.
+    const fakeProvider = {
+        getNetwork: async () => Network.from(1),
+        getTransactionCount: async () => 7,
+        estimateGas: async () => 21000n,
+        getFeeData: async () => ({
+            gasPrice: 2000000000n,
+            maxFeePerGas: 2000000000n,
+            maxPriorityFeePerGas: 1000000000n,
+        }),
+    };
+
+    // A private-key wallet as it is persisted in state, so the test goes
+    // through getSignerForAddress() the way the popup does.
+    const walletData = { type: "privkey" };
+
+    async function popupSignsTx(txParams) {
+        const localSigner = getSignerForAddress(walletData, 0, SIGNER_KEY);
+        const connected = localSigner.connect(fakeProvider);
+        const populated = await connected.populateTransaction(txParams);
+        delete populated.from;
+        return connected.signTransaction(populated);
+    }
+
+    test("a populated, signed transaction is accepted and broadcastable", async () => {
+        const rawSignedTx = await popupSignsTx(TX_PARAMS);
+        const parsed = verifySignedTx(rawSignedTx, TX_PARAMS, signer.address);
+        expect(parsed.nonce).toBe(7);
+        expect(parsed.chainId).toBe(1n);
+        expect(parsed.gasLimit).toBe(21000n);
+        expect(parsed.to).toBe(RECIPIENT);
+        expect(parsed.value).toBe(BigInt(TX_PARAMS.value));
+        expect(parsed.data).toBe(TX_PARAMS.data);
+        expect(parsed.signature).not.toBeNull();
+    });
+
+    test("the wire payload carries no password and no secret", async () => {
+        const rawSignedTx = await popupSignsTx(TX_PARAMS);
+        const payload = {
+            type: "AUTISTMASK_TX_RESPONSE",
+            id: "test-approval-id",
+            approved: true,
+            rawSignedTx,
+        };
+        expect(Object.keys(payload).sort()).toEqual([
+            "approved",
+            "id",
+            "rawSignedTx",
+            "type",
+        ]);
+        const wire = JSON.stringify(payload).toLowerCase();
+        expect(wire).not.toContain("password");
+        expect(wire).not.toContain(SIGNER_KEY.slice(2).toLowerCase());
+    });
+
+    test("the background rejects a transaction the popup did not approve", async () => {
+        const rawSignedTx = await popupSignsTx({
+            ...TX_PARAMS,
+            to: OTHER_RECIPIENT,
+        });
+        expect(() =>
+            verifySignedTx(rawSignedTx, TX_PARAMS, signer.address),
+        ).toThrow(/approved recipient/);
+    });
+});

tests/e2e/harness.js

@@ -0,0 +1,289 @@
+// End-to-end harness: launches a real Chromium with the unpacked MV3
+// build loaded, collects every uncaught page error and console.error, and
+// exposes the popup flows the tests drive.
+//
+// This runs inside the pinned Playwright container; see script/test-e2e.
+// It is deliberately NOT part of make check — REPO_POLICIES.md caps
+// make test at 20 seconds and a browser suite does not fit.
+
+"use strict";
+
+const fs = require("fs");
+const os = require("os");
+const path = require("path");
+
+const { chromium } = require("playwright-core");
+const { installNetworkStubs } = require("./network");
+
+const REPO_ROOT = path.resolve(__dirname, "..", "..");
+const EXT_PATH = path.join(REPO_ROOT, "dist", "chrome");
+
+// Page errors that are known, tracked, and deliberately tolerated. Every
+// entry must name the issue that will remove it. This list is the one
+// concession in an otherwise zero-tolerance policy: an uncaught error is
+// how this harness caught issue #150 in the first place.
+const ALLOWED_ERRORS = [
+    {
+        // libsodium ships a WASM build and an asm.js fallback. The
+        // extension CSP (script-src 'self', with no wasm-unsafe-eval)
+        // refuses the WASM module on every popup load; libsodium catches
+        // it and falls back to asm.js, so the wallet works. Deciding
+        // which backend actually ships is issue #182, and this entry gets
+        // deleted when that lands.
+        issue: "#182",
+        pattern: /Refused to compile or instantiate WebAssembly module/,
+    },
+];
+
+function isAllowed(text) {
+    return ALLOWED_ERRORS.some((a) => a.pattern.test(text));
+}
+
+// Collects every uncaught page error, console.error and unstubbed
+// request, and hands each one to exactly one reporter.
+//
+// This deliberately has NO window API. It used to expose mark()/since()
+// so a test could ask for "the errors since I started", and that shape
+// produced a green run that proved nothing twice over: first the mark
+// started after test 1, so everything recorded during launch was
+// discarded, then the tail after the final test was never read at all. In
+// both cases a record fell outside somebody's window and vanished, which
+// is the precise failure this harness exists to prevent.
+//
+// So there is no window left to fall outside of. take() is the only
+// reader and it always takes everything outstanding, so successive takes
+// partition the entire record stream with no gaps, and the runner turns
+// every record it reads into a failure.
+//
+// Observation ends when the browser context is closed. Nothing records
+// after that — the route handler and the console listeners are gone with
+// the context — so there is no post-teardown phase to collect, and this
+// class deliberately offers no mechanism pretending to cover one.
+class ErrorCollector {
+    constructor() {
+        this.entries = [];
+        this.taken = 0;
+    }
+
+    record(kind, text) {
+        const line = kind + ": " + String(text).split("\n")[0];
+        if (isAllowed(line)) return;
+        this.entries.push(line);
+    }
+
+    // Everything recorded since the previous take(). Never yields a
+    // record twice and never skips one.
+    take() {
+        const out = this.entries.slice(this.taken);
+        this.taken = this.entries.length;
+        return out;
+    }
+}
+
+function attachErrorListeners(ctx, errors) {
+    const attachPage = (page) => {
+        page.on("pageerror", (err) => {
+            errors.record("pageerror", err.message || String(err));
+        });
+        page.on("console", (msg) => {
+            if (msg.type() === "error") {
+                errors.record("console.error", msg.text());
+            }
+        });
+    };
+    ctx.pages().forEach(attachPage);
+    ctx.on("page", attachPage);
+    // Per-page listeners only: the context-level "weberror" event covers
+    // the same page exceptions and would double-report them. Playwright
+    // exposes no error EVENT for service workers, so an uncaught
+    // exception in the background worker is not visible here — everything
+    // this suite drives lives in the popup page. That is an error-channel
+    // gap only: worker NETWORK traffic is intercepted and reported like
+    // any other, and assertWorkerTrafficIntercepted() below fails the run
+    // if it ever stops being.
+}
+
+async function serviceWorker(ctx) {
+    const [existing] = ctx.serviceWorkers();
+    if (existing) return existing;
+    return ctx.waitForEvent("serviceworker", { timeout: 30000 });
+}
+
+// How long to wait for the background worker's first outbound request.
+//
+// The margin that actually decides whether this check is sound is not
+// this timeout — it is whether the route handler is installed before the
+// worker fetches. Measured over several runs: route installation
+// completes 11-23ms after the context comes up, and the worker's
+// blocklist fetch arrives 525-883ms after that, so the route wins by
+// roughly 25-50x. This 30s figure is only slack for a loaded machine on
+// top of that; losing the race fails the run rather than passing it
+// quietly, which was verified by forcing a 3s delay before route
+// installation.
+const WORKER_TRAFFIC_TIMEOUT_MS = 30000;
+
+// ctx.route() only sees service-worker requests when Playwright runs with
+// PW_EXPERIMENTAL_SERVICE_WORKER_NETWORK_EVENTS=1, which script/test-e2e
+// sets. Without it the worker's traffic — notably the phishing blocklist
+// fetch src/background/index.js issues at startup — goes to the real
+// internet, and nothing says so, because src/shared/phishingDomains.js
+// swallows fetch failures. A harness whose isolation can lapse in silence
+// is worthless, so this does not take the flag on trust: the background
+// worker's own startup fetch has to show up in the route handler, or the
+// suite refuses to run.
+//
+// Deliberately NOT a synthetic probe fetched through worker.evaluate():
+// evaluating in an extension worker this early kills it (the call fails
+// with "Target page, context or browser has been closed" and the worker
+// disappears), which would break the very thing being measured. Observing
+// traffic the extension already generates costs nothing and cannot
+// perturb it.
+async function assertWorkerTrafficIntercepted(stubs) {
+    const seen = await stubs.waitForServiceWorkerTraffic(
+        WORKER_TRAFFIC_TIMEOUT_MS,
+    );
+    if (seen) return seen;
+
+    // State the observation, not a conclusion. This fires for at least
+    // two quite different causes and the harness cannot tell them apart
+    // from here, so guessing one of them in the message sends the reader
+    // the wrong way.
+    throw new Error(
+        "observed no service-worker request in the route handler within " +
+            WORKER_TRAFFIC_TIMEOUT_MS +
+            "ms. Under working interception the background worker's " +
+            "startup blocklist fetch (src/background/index.js) reaches the " +
+            "handler about half a second after the route is installed. " +
+            "Two causes are plausible and this check cannot distinguish " +
+            "them: (1) service-worker interception is not in effect, so " +
+            "that traffic went to the real internet unobserved — the suite " +
+            "must be run through script/test-e2e, which sets " +
+            "PW_EXPERIMENTAL_SERVICE_WORKER_NETWORK_EVENTS=1, and a " +
+            "Playwright upgrade may have dropped or renamed that flag; " +
+            "(2) no worker request was made in the first place — the route " +
+            "lost the startup race, or the worker no longer fetches at " +
+            "startup, in which case this check needs a new anchor because " +
+            "there is no longer any worker traffic to observe. Either way " +
+            "the fix is a replacement mechanism or an honest downgrade of " +
+            "the isolation claims in tests/e2e/network.js and README.md — " +
+            "not deleting this check",
+    );
+}
+
+async function launch(routeOpts) {
+    if (!fs.existsSync(path.join(EXT_PATH, "manifest.json"))) {
+        throw new Error(
+            "no unpacked build at " +
+                EXT_PATH +
+                " — run make build before the e2e suite",
+        );
+    }
+
+    const userDir = fs.mkdtempSync(path.join(os.tmpdir(), "autistmask-e2e-"));
+    const ctx = await chromium.launchPersistentContext(userDir, {
+        // channel: "chromium" is load-bearing. The default headless mode
+        // uses the headless shell, which silently refuses to load
+        // extensions: there is no error at all, the service worker simply
+        // never appears. This cost real debugging time once already.
+        channel: "chromium",
+        headless: true,
+        args: [
+            "--disable-extensions-except=" + EXT_PATH,
+            "--load-extension=" + EXT_PATH,
+            // The container runs unprivileged; Chrome's sandbox needs
+            // capabilities the harness deliberately does not grant it.
+            "--no-sandbox",
+            // Belt to the interception braces: nothing that slips past
+            // the route handler can resolve a name, so a request that
+            // escapes cannot actually reach the internet. Detection is
+            // still assertWorkerTrafficIntercepted()'s job — this only
+            // bounds the damage while a gap goes unnoticed. Playwright
+            // fulfils routed requests without touching the resolver, and
+            // it drives the browser over a pipe, so neither is affected.
+            "--host-resolver-rules=MAP * ~NOTFOUND",
+        ],
+    });
+
+    const cleanup = async () => {
+        await ctx.close().catch(() => {});
+        fs.rmSync(userDir, { recursive: true, force: true });
+    };
+
+    try {
+        const errors = new ErrorCollector();
+        attachErrorListeners(ctx, errors);
+        routeOpts.report = (text) => errors.record("network", text);
+        const stubs = await installNetworkStubs(ctx, routeOpts);
+
+        await assertWorkerTrafficIntercepted(stubs);
+
+        // The extension id is derived from the unpacked path, so it
+        // changes and must never be hardcoded. It is the host part of the
+        // service worker URL.
+        const sw = await serviceWorker(ctx);
+        const id = new URL(sw.url()).host;
+
+        return {
+            ctx,
+            errors,
+            extensionId: id,
+            popupUrl: "chrome-extension://" + id + "/src/popup/index.html",
+            close: cleanup,
+        };
+    } catch (e) {
+        // Anything that fails after the browser is up has to tear it down
+        // on the way out: an orphaned context keeps node alive forever,
+        // turning a clean failure into a hung run.
+        await cleanup();
+        throw e;
+    }
+}
+
+// ---------------------------------------------------------------- flows
+
+const PASSWORD = "e2e-harness-password";
+
+async function visible(page, selector, timeout = 15000) {
+    await page.waitForSelector(selector, { state: "visible", timeout });
+}
+
+async function openPopup(ctx, popupUrl) {
+    const page = await ctx.newPage();
+    await page.goto(popupUrl);
+    return page;
+}
+
+// Full wallet creation through the real UI: BIP-39 generation, libsodium
+// vault encryption and extension storage persistence, for real.
+async function createWallet(page) {
+    await page.click("#btn-welcome-add");
+    await visible(page, "#view-add-wallet");
+    await page.click("#btn-generate-phrase");
+    await page.waitForFunction(() => {
+        const el = document.getElementById("wallet-mnemonic");
+        return el && el.value.trim().split(/\s+/).length >= 12;
+    });
+    await page.fill("#add-wallet-password", PASSWORD);
+    await page.fill("#add-wallet-password-confirm", PASSWORD);
+    await page.click("#btn-add-wallet-confirm");
+    await visible(page, "#view-main", 60000);
+}
+
+// Reach the address detail screen from wherever the popup restored to.
+// Clicking .address-row does not open it; the [info] button does.
+async function openAddressDetail(page) {
+    const onAddress = await page.isVisible("#view-address");
+    if (!onAddress) {
+        await visible(page, "#view-main");
+        await page.click("#wallet-list .btn-addr-info");
+    }
+    await visible(page, "#view-address");
+}
+
+module.exports = {
+    createWallet,
+    launch,
+    openAddressDetail,
+    openPopup,
+    visible,
+};

tests/e2e/network.js

@@ -0,0 +1,334 @@
+// Browser-level network interception for the end-to-end suite.
+//
+// Every http(s) request the extension makes — from the popup page AND
+// from the MV3 background service worker — is fulfilled from these
+// fixtures, so the suite is deterministic and runs entirely offline. The
+// probe that motivated this harness (see issue #181) observed live calls
+// to Blockscout returning 401 inside the container, which would make any
+// assertion about rendered transaction data worthless.
+//
+// Service-worker coverage is not free: ctx.route() only sees worker
+// traffic when PW_EXPERIMENTAL_SERVICE_WORKER_NETWORK_EVENTS=1 is set in
+// the environment, which script/test-e2e does. Without it the phishing
+// blocklist fetch that src/background/index.js issues at worker startup
+// silently reaches raw.githubusercontent.com on the open internet, and
+// src/shared/phishingDomains.js swallows the failure so nothing surfaces
+// it. That is not left to trust: waitForServiceWorkerTraffic() below
+// backs the launch-time canary in harness.js, which fails the entire
+// suite if worker requests stop being visible here.
+//
+// Anything not explicitly stubbed here is aborted AND reported to the
+// error collector, so a newly added outbound call shows up as a test
+// failure rather than as intermittent flakiness.
+
+"use strict";
+
+// Fictional ERC-20 used to seed the transaction-detail test. The symbol
+// must not collide with any entry in src/shared/tokenList.js, or
+// isSpoofedSymbol() in src/shared/transactions.js drops the transfer as a
+// symbol-spoofing attempt; holders_count must be >= 1000 or the default
+// hideLowHolderTokens filter drops it. Either would make the test pass
+// vacuously by never rendering a row at all.
+const STUB_TOKEN = {
+    address: "0xe2e0000000000000000000000000000000000e2e",
+    symbol: "E2E",
+    name: "End To End Test Token",
+    decimals: "6",
+    holders: "12345",
+};
+
+const STUB_COUNTERPARTY = "0xc0ffee0000000000000000000000000000c0ffee";
+
+const STUB_TX_HASH =
+    "0xe2e0000000000000000000000000000000000000000000000000000000000e2e";
+
+const STUB_BLOCK_NUMBER = 21000000;
+
+// Fixed instant so timeAgo() output is stable across runs.
+const STUB_TX_TIMESTAMP = "2026-01-02T03:04:05.000000Z";
+
+// A 32-byte zero word. Returned for every eth_call, which is what makes
+// ethers' ENS reverse lookup resolve to "no resolver set" and return null
+// instead of throwing. A throw would be logged by src/shared/ens.js via
+// log.errorf(), i.e. console.error, which fails the run on its own.
+const ZERO_WORD = "0x" + "0".repeat(64);
+
+const RPC_RESULTS = {
+    eth_chainId: "0x1",
+    net_version: "1",
+    eth_blockNumber: "0x1406f40",
+    eth_getBalance: "0x0",
+    eth_call: ZERO_WORD,
+    eth_gasPrice: "0x3b9aca00",
+    eth_estimateGas: "0x5208",
+    eth_getTransactionCount: "0x0",
+    eth_maxPriorityFeePerGas: "0x3b9aca00",
+};
+
+function tokenObject() {
+    return {
+        address_hash: STUB_TOKEN.address,
+        address: STUB_TOKEN.address,
+        symbol: STUB_TOKEN.symbol,
+        name: STUB_TOKEN.name,
+        decimals: STUB_TOKEN.decimals,
+        holders_count: STUB_TOKEN.holders,
+        type: "ERC-20",
+    };
+}
+
+// One received ERC-20 transfer of 1.5 E2E to the address under test.
+function tokenTransferItems(address) {
+    return [
+        {
+            transaction_hash: STUB_TX_HASH,
+            block_number: STUB_BLOCK_NUMBER,
+            timestamp: STUB_TX_TIMESTAMP,
+            from: { hash: STUB_COUNTERPARTY },
+            to: { hash: address },
+            total: { decimals: STUB_TOKEN.decimals, value: "1500000" },
+            token: tokenObject(),
+        },
+    ];
+}
+
+// Full details for STUB_TX_HASH. raw_input is "0x" so the calldata
+// decoder short-circuits; the on-chain detail fields still populate.
+function transactionDetails() {
+    return {
+        hash: STUB_TX_HASH,
+        block_number: STUB_BLOCK_NUMBER,
+        nonce: 7,
+        gas_used: "51000",
+        gas_price: "1000000000",
+        fee: { value: "51000000000000" },
+        raw_input: "0x",
+        status: "ok",
+    };
+}
+
+function jsonResponse(route, body) {
+    return route.fulfill({
+        status: 200,
+        contentType: "application/json",
+        body: JSON.stringify(body),
+    });
+}
+
+// Extract the address from a Blockscout /addresses/<addr>/... path.
+function blockscoutAddress(pathname) {
+    const m = pathname.match(/\/addresses\/(0x[0-9a-fA-F]{40})\//);
+    return m ? m[1] : null;
+}
+
+function handleRpc(route, postData, report) {
+    let payload;
+    try {
+        payload = JSON.parse(postData || "null");
+    } catch {
+        report("unstubbed RPC: unparseable body " + String(postData));
+        return route.abort();
+    }
+    // ethers batches by default, so the body may be an array.
+    const batch = Array.isArray(payload) ? payload : [payload];
+
+    // Anything that is not a JSON-RPC object, or a batch of them, is not
+    // RPC at all and must be reported like any other unrecognised
+    // outbound traffic rather than dereferenced. request.postData()
+    // returns null both for a bodyless POST and for a body Playwright
+    // cannot decode as UTF-8 (sendBeacon with a Blob, or any binary
+    // payload), so this is not an empty-string special case: it rejects
+    // every non-object payload, exactly as the catch above rejects every
+    // unparseable one.
+    if (
+        payload === null ||
+        typeof payload !== "object" ||
+        !batch.every((req) => req !== null && typeof req === "object")
+    ) {
+        report("unstubbed request: POST " + route.request().url());
+        return route.abort();
+    }
+    const replies = batch.map((req) => {
+        const result = RPC_RESULTS[req.method];
+        if (result === undefined) {
+            report("unstubbed RPC method: " + req.method);
+            return {
+                jsonrpc: "2.0",
+                id: req.id,
+                error: { code: -32601, message: "unstubbed in e2e harness" },
+            };
+        }
+        return { jsonrpc: "2.0", id: req.id, result };
+    });
+    return jsonResponse(route, Array.isArray(payload) ? replies : replies[0]);
+}
+
+const TRACE_TRUE = ["1", "true", "yes", "on"];
+const TRACE_FALSE = ["", "0", "false", "no", "off"];
+
+// Whether E2E_TRACE_NETWORK asks for the request trace.
+//
+// A set-but-unrecognised value is a hard error rather than a quiet
+// "off": E2E_TRACE_NETWORK=true asking for a trace and getting silence
+// is the operator being lied to about what the harness is doing, which
+// is the whole failure mode this suite exists to eliminate. Refusing to
+// guess costs one line and one obvious error message.
+function traceEnabled(raw) {
+    if (raw === undefined || raw === null) return false;
+    const v = String(raw).trim().toLowerCase();
+    if (TRACE_TRUE.includes(v)) return true;
+    if (TRACE_FALSE.includes(v)) return false;
+    throw new Error(
+        "E2E_TRACE_NETWORK is set to " +
+            JSON.stringify(String(raw)) +
+            ", which is not a recognised on/off value. Use one of " +
+            TRACE_TRUE.join(", ") +
+            " to enable the request trace, or one of " +
+            TRACE_FALSE.slice(1).join(", ") +
+            " to disable it. Refusing to guess: a diagnostic that silently " +
+            "does nothing is worse than one that is not there",
+    );
+}
+
+/**
+ * Route every http(s) request through local fixtures.
+ *
+ * @param {import("playwright-core").BrowserContext} ctx
+ * @param {object} opts
+ * @param {(text: string) => void} opts.report  called for unstubbed traffic
+ * @param {boolean} [opts.seedTokenTransfer]  serve the stubbed ERC-20
+ *   transfer. Read at request time, so a test can flip it on the same
+ *   options object without re-registering the route.
+ * @returns {Promise<{waitForServiceWorkerTraffic: (ms: number) =>
+ *   Promise<string|null>}>}
+ */
+async function installNetworkStubs(ctx, opts) {
+    const report = opts.report;
+
+    // First request seen that originated in a service worker, and the
+    // resolver waiting for it. This is what proves worker interception is
+    // actually in force; see waitForServiceWorkerTraffic below.
+    let firstWorkerRequest = null;
+    let announceWorkerRequest = null;
+
+    // E2E_TRACE_NETWORK=1 prints every request that reaches this handler,
+    // tagged [sw] when it originated in the background service worker.
+    // It exists so the isolation claim above can be re-checked by anyone
+    // in one command, without editing files: the phishing blocklist fetch
+    // showing up with an [sw] tag is the proof that the worker really is
+    // intercepted and that the raw.githubusercontent.com stub below is
+    // live code rather than decoration.
+    const trace = traceEnabled(process.env.E2E_TRACE_NETWORK);
+
+    // Regex rather than a glob so chrome-extension:// resource loads are
+    // never touched — routing those would break the popup itself.
+    await ctx.route(/^https?:\/\//, async (route) => {
+        const req = route.request();
+        const url = new URL(req.url());
+        const p = url.pathname;
+        const fromWorker = !!req.serviceWorker();
+
+        if (fromWorker && !firstWorkerRequest) {
+            firstWorkerRequest = req.method() + " " + req.url();
+            if (announceWorkerRequest)
+                announceWorkerRequest(firstWorkerRequest);
+        }
+
+        if (trace) {
+            const origin = fromWorker ? "[sw] " : "[page] ";
+            console.log("# routed " + origin + req.method() + " " + req.url());
+        }
+
+        // JSON-RPC endpoint (any host): a POST with a JSON-RPC body.
+        if (req.method() === "POST") {
+            return handleRpc(route, req.postData(), report);
+        }
+
+        // Blockscout v2
+        if (p.includes("/api/v2/")) {
+            if (/\/addresses\/0x[0-9a-fA-F]{40}\/transactions$/.test(p)) {
+                return jsonResponse(route, { items: [] });
+            }
+            if (/\/addresses\/0x[0-9a-fA-F]{40}\/token-transfers$/.test(p)) {
+                const addr = blockscoutAddress(p);
+                return jsonResponse(route, {
+                    items:
+                        opts.seedTokenTransfer && addr
+                            ? tokenTransferItems(addr)
+                            : [],
+                });
+            }
+            if (/\/addresses\/0x[0-9a-fA-F]{40}\/token-balances$/.test(p)) {
+                return jsonResponse(route, []);
+            }
+            if (p.endsWith("/transactions/" + STUB_TX_HASH)) {
+                return jsonResponse(route, transactionDetails());
+            }
+        }
+
+        // CoinDesk price tick
+        if (url.hostname.endsWith("coindesk.com")) {
+            return jsonResponse(route, { Data: {} });
+        }
+
+        // MetaMask phishing blocklist
+        if (
+            url.hostname === "raw.githubusercontent.com" ||
+            p.endsWith("/eth-phishing-detect/main/src/config.json")
+        ) {
+            return jsonResponse(route, {
+                version: 2,
+                tolerance: 2,
+                fuzzylist: [],
+                whitelist: [],
+                blacklist: [],
+            });
+        }
+
+        // Best-effort Etherscan address labels: served as an empty page.
+        if (url.hostname.endsWith("etherscan.io")) {
+            return route.fulfill({
+                status: 200,
+                contentType: "text/html",
+                body: "<html><body></body></html>",
+            });
+        }
+
+        report("unstubbed request: " + req.method() + " " + req.url());
+        return route.abort();
+    });
+
+    return {
+        /**
+         * Resolve with the first service-worker-originated request this
+         * handler saw, or null if none arrives within `ms`.
+         *
+         * The background worker fetches the phishing blocklist at
+         * startup, unconditionally, within about a second of the context
+         * coming up — so under working interception this resolves almost
+         * immediately. Nothing arriving means worker traffic is bypassing
+         * the handler entirely and going to the real internet, which the
+         * caller turns into a hard failure of the whole suite.
+         */
+        waitForServiceWorkerTraffic(ms) {
+            if (firstWorkerRequest) return Promise.resolve(firstWorkerRequest);
+            return new Promise((resolve) => {
+                const timer = setTimeout(() => {
+                    announceWorkerRequest = null;
+                    resolve(null);
+                }, ms);
+                announceWorkerRequest = (req) => {
+                    clearTimeout(timer);
+                    announceWorkerRequest = null;
+                    resolve(req);
+                };
+            });
+        },
+    };
+}
+
+module.exports = {
+    installNetworkStubs,
+    STUB_TOKEN,
+    STUB_TX_HASH,
+};

tests/e2e/run.js

@@ -0,0 +1,264 @@
+// End-to-end suite entrypoint. Run via script/test-e2e (which builds
+// dist/chrome/ and starts the pinned container); running it directly
+// requires a Chromium that playwright-core can find.
+//
+// A plain runner rather than jest on purpose: jest's default testMatch
+// would pull these files into script/test, and browser tests do not fit
+// inside the 20-second cap REPO_POLICIES.md puts on make test. Nothing
+// here is named *.test.js for the same reason.
+
+"use strict";
+
+const {
+    createWallet,
+    launch,
+    openAddressDetail,
+    openPopup,
+    visible,
+} = require("./harness");
+const { STUB_TOKEN, STUB_TX_HASH } = require("./network");
+
+const TEST_TIMEOUT_MS = 120000;
+
+// How long to keep collecting after the final test returns; see the
+// trailing drain in main().
+const TRAILING_WATCH_MS = 1500;
+
+const tests = [];
+
+function test(name, fn) {
+    tests.push({ name, fn });
+}
+
+function assert(cond, message) {
+    if (!cond) throw new Error(message);
+}
+
+function withTimeout(promise, name) {
+    let timer;
+    const timeout = new Promise((_, reject) => {
+        timer = setTimeout(
+            () =>
+                reject(new Error("timed out after " + TEST_TIMEOUT_MS + "ms")),
+            TEST_TIMEOUT_MS,
+        );
+    });
+    return Promise.race([promise, timeout]).finally(() => clearTimeout(timer));
+}
+
+// ----------------------------------------------------------------- tests
+
+test("popup loads and reaches the welcome view", async (env) => {
+    env.page = await openPopup(env.ctx, env.popupUrl);
+    await visible(env.page, "#view-welcome");
+    const title = await env.page.title();
+    assert(title === "AutistMask", "unexpected popup title: " + title);
+});
+
+test("wallet creation through the UI reaches the main view", async (env) => {
+    await createWallet(env.page);
+    const addrCount = await env.page
+        .locator("#wallet-list .btn-addr-info")
+        .count();
+    assert(addrCount > 0, "no addresses rendered in the wallet list");
+});
+
+test("add token screen opens from address detail (#150)", async (env) => {
+    await openAddressDetail(env.page);
+    await env.page.click("#btn-add-token");
+    await visible(env.page, "#view-add-token");
+    const quickPicks = await env.page
+        .locator("#common-token-list .common-token")
+        .count();
+    assert(quickPicks > 0, "no common-token quick-pick buttons rendered");
+});
+
+test("transaction detail renders an ERC-20 transfer (#151)", async (env) => {
+    // Serve the stubbed token transfer from here on, then reload so the
+    // address detail screen refetches its transaction list.
+    env.routeOpts.seedTokenTransfer = true;
+    await env.page.reload();
+    await openAddressDetail(env.page);
+
+    await visible(env.page, "#tx-list .tx-row");
+    const rowText = await env.page
+        .locator("#tx-list .tx-row")
+        .first()
+        .innerText();
+    assert(
+        rowText.includes(STUB_TOKEN.symbol),
+        "token transfer row missing symbol " +
+            STUB_TOKEN.symbol +
+            ", got: " +
+            JSON.stringify(rowText),
+    );
+
+    await env.page.locator("#tx-list .tx-row").first().click();
+    await visible(env.page, "#view-transaction");
+
+    const hash = await env.page.locator("#tx-detail-hash").innerText();
+    assert(
+        hash.includes(STUB_TX_HASH),
+        "transaction detail shows the wrong hash: " + hash,
+    );
+
+    // The token contract row is the field that crashes when
+    // addressDotHtml is not imported: it renders only for transfers with
+    // a contractAddress, which is every ERC-20 transfer.
+    await visible(env.page, "#tx-detail-token-contract-section");
+    const contract = env.page.locator("#tx-detail-token-contract");
+    const contractText = await contract.innerText();
+    assert(
+        contractText.toLowerCase().includes(STUB_TOKEN.address),
+        "token contract row missing the contract address, got: " +
+            JSON.stringify(contractText),
+    );
+    const dots = await contract.locator('span[style*="border-radius"]').count();
+    assert(dots > 0, "token contract row rendered without its colour dot");
+});
+
+// ---------------------------------------------------------------- runner
+
+async function main() {
+    // A suite that runs nothing must never report success. If a refactor
+    // drops the registrations above, or a require() of this file stops
+    // reaching them, the only honest outcome is a red run — reporting
+    // "0/0 passed" and exiting 0 is the same vacuous-check failure this
+    // whole harness exists to prevent.
+    if (tests.length === 0) {
+        console.log("1..0");
+        console.log("# FAILED: the e2e suite registered no tests");
+        process.exitCode = 1;
+        return;
+    }
+
+    const routeOpts = { seedTokenTransfer: false };
+
+    let session;
+    try {
+        session = await launch(routeOpts);
+    } catch (e) {
+        // Never skip and report success: a browser we cannot start, or
+        // one whose network interception is not in force, is a failure of
+        // the suite, not an absent one.
+        console.error("e2e: cannot run the suite: " + e.message);
+        process.exitCode = 1;
+        return;
+    }
+
+    console.log("# extension id: " + session.extensionId);
+    console.log("1.." + tests.length);
+
+    const env = {
+        ctx: session.ctx,
+        popupUrl: session.popupUrl,
+        routeOpts,
+        page: null,
+    };
+
+    // Attribution of collected errors is total. session.errors has no
+    // window API at all: take() always drains everything outstanding, so
+    // successive takes partition the whole stream, and the phases below
+    // cover the entire life of the run. Nothing the collector holds can
+    // go unread.
+    //
+    //   launch .. end of test 1     -> test 1 (so the worker's startup
+    //                                  fetches land on a test, not
+    //                                  nowhere)
+    //   end of test k .. end of k+1 -> test k+1
+    //   last test .. teardown       -> the suite, via the trailing drain
+    //
+    // Those three phases cover the entire life of the browser context.
+    // There is no fourth: once the context is closed nothing can record,
+    // because the route handler and the console listeners died with it.
+    // Traffic that a test defers past the trailing drain is therefore
+    // never observed at all — a real limit of this design, stated in the
+    // README, and not one any post-teardown hook could close.
+    //
+    // Two green-but-vacuous runs on this harness were the same shape: a
+    // record falling outside somebody's window and being dropped. First
+    // the mark started after test 1, discarding launch-time records;
+    // then the tail after the last test was never read. Patching a
+    // second boundary would have invited a third, so the window concept
+    // is gone rather than fixed.
+    let failed = 0;
+    let n = 0;
+    for (const t of tests) {
+        n += 1;
+        let failure = null;
+        try {
+            await withTimeout(t.fn(env), t.name);
+        } catch (e) {
+            failure = e.message;
+        }
+
+        // Any uncaught page error, console.error or unstubbed request
+        // fails the test that provoked it, whether or not its assertions
+        // passed. This is the mechanism that caught #150.
+        const newErrors = session.errors.take();
+        if (!failure && newErrors.length > 0) {
+            failure = "uncaught browser errors during this test";
+        }
+
+        if (failure) {
+            failed += 1;
+            console.log("not ok " + n + " - " + t.name);
+            console.log("  " + failure);
+            for (const line of newErrors) {
+                console.log("  " + line);
+            }
+        } else {
+            console.log("ok " + n + " - " + t.name);
+        }
+    }
+
+    // Keep watching after the last test returns, before tearing the
+    // browser down. A request a test fires without awaiting is still in
+    // flight when its function resolves; measured here it reaches the
+    // route handler about 10ms later, but closing the context does not
+    // wait for it — with no window at all the request dies unobserved
+    // and the run goes green, which is exactly how escaping traffic
+    // stays invisible.
+    //
+    // A fixed bounded window rather than a quiescence poll on purpose:
+    // the collector being quiet is not evidence, because a request that
+    // has not been dispatched yet has recorded nothing to be quiet
+    // about. Playwright offers no "is anything in flight" question to
+    // ask either — the route handler is the only observation point — so
+    // a grace period is the mechanism available, and this one is ~150x
+    // the measured latency for 1.5s on a ~25s suite.
+    await new Promise((resolve) => setTimeout(resolve, TRAILING_WATCH_MS));
+
+    await session.close();
+
+    // The tail. These cannot be blamed on any single test, so they are
+    // reported against the suite rather than guessed at — but they are
+    // reported, and they fail the run.
+    const trailing = session.errors.take();
+
+    console.log(
+        "# " + (tests.length - failed) + "/" + tests.length + " tests passed",
+    );
+
+    if (trailing.length > 0) {
+        console.log(
+            "# " +
+                trailing.length +
+                " browser error(s) recorded after the last test finished, " +
+                "not attributable to any single test:",
+        );
+        for (const line of trailing) {
+            console.log("#   " + line);
+        }
+    }
+
+    if (failed > 0 || trailing.length > 0) {
+        console.log("# FAILED");
+        process.exitCode = 1;
+    }
+}
+
+main().catch((e) => {
+    console.error("e2e: " + (e && e.stack ? e.stack : e));
+    process.exitCode = 1;
+});

yarn.lock

@@ -2547,6 +2547,11 @@ pkg-dir@^4.2.0:
   dependencies:
     find-up "^4.0.0"
 
+playwright-core@1.56.0:
+  version "1.56.0"
+  resolved "https://registry.yarnpkg.com/playwright-core/-/playwright-core-1.56.0.tgz#14b40ea436551b0bcefe19c5bfb8d1804c83739c"
+  integrity sha512-1SXl7pMfemAMSDn5rkPeZljxOCYAmQnYLBTExuh6E8USHXGSX3dx6lYZN/xPpTz1vimXmPA9CDnILvmJaB8aSQ==
+
 pngjs@^5.0.0:
   version "5.0.0"
   resolved "https://registry.npmjs.org/pngjs/-/pngjs-5.0.0.tgz"