The blocklist URL in shipped code named a competitor and pointed at a moving
ref, and the extension re-fetched from it every 24 hours, which also meant a
third party decided what this wallet warns about. All of that is gone.
script/vendor-blocklist fetches upstream at a pinned commit, verifies the
sha256 of the bytes that commit serves, and writes
src/shared/phishingBlocklist.json. It is build-time tooling, never shipped, and
the one place in the repo that names the upstream project; a source reference
nobody can verify is not a source reference.
The artifact stores truncated sha256 digests rather than domain names. That is
what censors it: the previous file contained the competitor's name 6,475 times,
as phishing domains impersonating them, and not one of those domains is
dropped. It also makes lookups a binary search over a fixed-width string, so
nothing is built at module load — which matters on MV3, where the worker
re-evaluates the module on every wake — and takes the file from 8.7 MB to
1.7 MB.
script/check-censored enforces the rest: it reads the name out of the vendoring
script rather than repeating it, and fails on any occurrence in the working
tree or under dist/ that is not one of the three literals shipped code cannot
avoid — two provider-shim identifiers in src/content/inpage.js and one ERC-20's
on-chain name in src/shared/tokenList.js. Each is permitted only at the path
that carries it, and at the emitted paths that path is bundled into, so a
literal appearing anywhere else fails like any other occurrence. It runs in
make check, which inspects dist/ when there is one and says loudly when there
is not, and again with --require-dist at the end of every make build.
Removing the runtime fetch retires the delta, the extension-storage persistence
and the 24-hour alarm from #158. A retired alarm is now cleared rather than
left waking the worker forever on installs that already have it.
The e2e suite drives the warning end to end from a real blocklisted origin
served as a real http(s) site, with a control asserting the banner stays hidden
for one that is not listed. Its service-worker interception canary needed a new
anchor, since the startup fetch it used to watch for no longer happens: it now
wakes the worker with a message and asks it for one throwaway fetch.
LICENSE no longer cites a repository that returns 404.
eslint.config.js gains one block: script/lib/ holds node programs the shell
entrypoints call, and without it they lint with no globals at all.
Seven bundled tokens were filtered as spoofs at their own address, so a user
holding FRAX, TON, REUSD, EURE, MSUSD, MUSD or JPYC could not see or spend the
one the wallet happened not to pick.
The known-symbol table is derived from the bundled token list, first-wins in
market-cap order, so a symbol that appears twice silently condemned its second
contract. Both are real tokens from the same fetch and neither is stale --
three pairs are one issuer's old and new contract, four are unrelated issuers
sharing a ticker. Picking a winner would have been guessing, and dropping the
ambiguous symbols would have ended spoof filtering for those tickers entirely.
The table now maps a symbol to the set of addresses that legitimately bear it.
A contract outside the set is still a spoof, so the check is not weakened: a
third contract bearing any of the seven shared tickers is refused, and that is
tested. The filter decides what is fake, not what is worth holding, so a legacy
contract stays in the set -- it still holds real balances.
A test walks the whole bundled list asserting no token is filtered at its own
address, which is the guard whose absence let this ship.
A token calling itself " ETH " missed the known-symbol table entirely, so the
spoof check reported it was not a spoof -- while HTML collapsed the whitespace
and displayed it as ETH next to the user's real ETH. One space defeated the
filter.
The symbol is now folded before the lookup: NFKC, remove what paints nothing,
trim, uppercase. The rule is "remove what paints nothing"; the Unicode classes
are how that is spelled, which is why U+007F is named separately -- it is a
control, reached by no class, and measures identical to no character at all.
Every width in the module comment was measured in the pinned browser rather
than reasoned about, and the boundary is pinned from both sides: widening to
all control characters fails the visible-controls test, narrowing back fails
the invisible-characters test. Two default-ignorable code points do paint a
box and are folded anyway, which can only hide a token that does not resemble
the symbol it folds to -- the harmless direction, recorded rather than glossed.
Confusables that are distinct letters, bidi reordering and interior whitespace
are knowingly left open and asserted open by tests.
