Critical - Any pubkey could publish a kind:30078 roles document and become root of a deployment, granting itself every capability and denying anyone. Roles are now accepted only from the configured root, compared against the constructor value rather than current authority state so a bad ingest cannot move the goalposts. Reproduced end to end before the fix; regression test added. - Administrative state now fails closed without signature verification. Without it, event.pubkey is an unauthenticated claim, so an authorship check alone proves nothing — a relay can put the root's key on an event the root never wrote. trustUnsignedEvents opts in for local and test use. High - Report categories could evade moderation entirely: resolveThresholds used bracket truthiness, so "constructor" resolved to the Object constructor (all gates undefined) and "__proto__" to Object.prototype, both bypassing the default threshold. Own-property lookups and null-prototype tables throughout core — policy, evaluator, evidence, authority, presets. - Live subscriptions bypassed signature verification entirely; an attacker only had to wait for the subscription rather than the load. - The default administrative query asked relays for every kind:30078 and kind:30000 event in existence, with no author, identifier, or limit. Now filtered to governance identifiers with a bounded limit. Medium - Cached state was trusted as authority: a cache naming a different root is now rejected, so same-origin tampering cannot persist a takeover. - Provider attribute changes silently stopped working after the first one, and concurrent starts leaked a transport. Generation-guarded. - Unbounded growth in the report store, mute store, decision cache, and the transport's per-subscription dedupe set — all now capped. Also restores the on-wire `d` identifier to <namespace>:governance:<scope>:v1. The BitGate rename replaced "governance:" too broadly and silently changed the event format; both encoder and parser changed together, so round-trip tests never noticed. The identifier names the kind of document, not the library. Adds docs/security.md and 21 regression tests reproducing each finding. 832 tests.
@bitgate/widget
Drop-in custom elements. Framework-agnostic, no bundler required by consumers, and built strictly on the published API.
Install
npm install && npm run build:widget
Copy packages/widget/dist/bitgate.js (and its .map) onto your own origin.
The bundle is deliberately not committed: build it from reviewed source rather
than trusting a prebuilt minified blob for a moderation widget.
The one-element version
<script type="module" src="/vendor/bitgate/bitgate.js"></script>
<bitgate-provider relays="wss://relay.example" root="<root-pubkey>" policy="social">
<bitgate-veil profile="feed" target-user="<author-pubkey>">
<img src="thumbnail.jpg" alt="" />
</bitgate-veil>
</bitgate-provider>
<bitgate-provider> builds the runtime, loads administrative state, and shares
both with every descendant. Nothing else to wire.
Provider attributes
| Attribute | Meaning |
|---|---|
relays |
Comma-separated relay URLs (required) |
root |
Root administrator pubkey, from your deployment config |
policy |
social, commerce, or admin-only (default) |
viewer |
Viewer pubkey, if known at render time |
application / namespace |
Storage and event namespacing |
Target attributes
Any target-bearing element reads its target from markup:
<bitgate-veil target-user="<pubkey>">…</bitgate-veil>
<bitgate-veil target-event="<id>" target-author="<pubkey>">…</bitgate-veil>
<bitgate-veil target-address="30078:<pubkey>:sku-001">…</bitgate-veil>
Assigning .target in JavaScript still works and takes precedence.
With BitLogin
const provider = document.querySelector("bitgate-provider");
await provider.ready;
await provider.useSigner(window.nostr);
Await ready rather than listening for bitgate:ready — the event fires during
element upgrade, which is before a script that runs after parsing could attach a
listener.
The rule these elements follow
They render decisions and issue commands. They compute no policy.
If an element here ever needed to compare a count to a threshold, that would mean the policy definition is missing a profile — not that the widget needs a number. Keeping that line is the whole reason the engine was extracted.
Elements
Viewer-facing
| Element | Purpose |
|---|---|
<bitgate-veil> |
Wraps content and applies the decision: blur on restrict, disclosure on hide, warning on warn |
<bitgate-report> |
Report dialog. Requires no capability — anyone may report |
<bitgate-status> |
Readout of effect, reasons, and evidence |
Moderator-facing
| Element | Purpose |
|---|---|
<bitgate-capabilities> |
What the signed-in account may actually do |
<bitgate-action> |
A capability-gated button that explains refusals |
<bitgate-admin-panel> |
Effective state, contributor attribution, and controls |
Usage
<script type="module">
import { createBitGate, createCommands } from "@bitgate/runtime";
import { defineBitGateElements } from "@bitgate/widget";
defineBitGateElements();
const runtime = createBitGate({
applicationId: "my-app",
namespace: "myapp",
root: ROOT_PUBKEY,
transport: myRelayAdapter,
policy: myPolicy,
now: () => Math.floor(Date.now() / 1000),
});
await runtime.loadAdministrativeState();
const veil = document.querySelector("bitgate-veil");
veil.runtime = runtime;
veil.target = { type: "event", id: eventId, author: authorPubkey };
</script>
<bitgate-veil profile="feed">
<img src="thumbnail.jpg" alt="" />
</bitgate-veil>
Elements receive a runtime; they never construct one. A widget that owned relay connections would make it impossible to share one governance runtime across a page.
Events
All events are composed so they cross shadow boundaries.
| Event | Fired when |
|---|---|
bitgate:revealed |
The viewer chose to show hidden content |
bitgate:reported / bitgate:report-failed |
A report was published or refused |
bitgate:action-completed / :action-refused / :action-failed |
An action resolved |
bitgate:account-restricted |
The admin panel restricted an account |
Wording
The engine emits stable reason identifiers, never prose, so applications phrase them for their own audience:
veil.reasonText = {
"admin-user-deny": "This seller is suspended",
"trusted-report-threshold": "Flagged by several buyers you follow",
};
Design decisions worth knowing
Hidden content is removed from the accessibility tree, not merely covered. A screen reader must not read out something the viewer was not meant to see.
A hidden decision renders a disclosure, not nothing. Content that silently vanishes is indistinguishable from a bug and leaves no route to appeal. Where the profile permits an override, a "Show anyway" control appears; where it does not, the reason still does.
Refused actions stay visible and explain themselves. A moderator who cannot see that they lack a capability will assume the tool is broken, and a hidden button teaches nothing about the authority model. The command layer enforces the same check independently — the UI is an explanation, never the gate.
Styling is yours. Elements expose part attributes and inherit color and
font from the host. Override the CSS custom properties (--gov-fg,
--gov-surface, --gov-warn, --gov-danger, --gov-radius) or target parts:
bitgate-veil::part(veil) { border: 1px solid red; }
Environment
defineBitGateElements() is safe to call repeatedly and safe to import in
Node — registration is a no-op without a DOM, so a server-rendering host can
import the helpers without guarding.