An ancestor's guarded-wallet list shows @alice with nothing saying which
group in the lineage signed her wallet's certificate, so two wallets
under different branches read the same. The plan names the direct
guardian (the certificate's signer, guardians[0]) by its founding name,
carries that group's birth certificate on the forwarding notice so a
device outside it can name it, and gathers a person's wallets under one
header on the group's wallet screen and one row per wallet on their
profile.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
member-wallet-delay.md is marked built, with a table of where the build
chose differently -- most of all that a request carrying no policy still
reads, so wallets guarded before policies keep verifying. The member
wallets note's "full control" stops saying 12, and its D3 adds that only
an open group takes a request; the guardian-delay note's member-wallet row
and the README follow.
The three-emulator run is still to do by hand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The member-wallet policy (30337) becomes the group's opt-in: no policy
means the group does not guard members' wallets (D8), so every new member
wallet's delay is one its group chose. A newer policy with ["closed"]
instead of a delay stops new wallets (D9); it never releases one already
guarded, whose addresses carry the group's leaf. Opening again is a newer
open policy.
The signer table, the picker's refusals (POLICY_UNKNOWN, NOT_OFFERED,
CLOSED), the request, the screen -- now MemberWalletPolicyScreen, with
Stop guarding new wallets -- and the emulator run follow. Phases 4, 5 and
6 now ship together, since 4 and 5 close every group until 6 opens it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The group's quorum signs a standing policy in its own room -- a new kind,
30337, ["guardian_delay", <blocks>] -- rather than a birth certificate
tag: the certificate is the parent's statement, frozen at birth, and a
root group has none. A member's request is built under the newest policy
and carries it whole, so late members (through the key notice) and
signers who missed it judge it alike; signers refuse a stale or missing
one. A wallet already guarded keeps its delay. Seven phases, not built.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brings in member guarded wallets and funding a descendant's guarded wallet.
- Schema: bennu's v30 (MemberWallet, MemberWalletSpend) is kept as it is;
DkgSession.guardianDelay moves to v31, with a 30 -> 31 auto-migration.
- Conflicts: strings and README take both sides; the send form's guardian
delay line moves into the shared SpendForm; the guarded wallet card keeps
its Fund and Spend buttons and gains this group's own leaf's delay.
- bennu's uses of ANCESTOR_DELAY become each leaf's own delay:
GuardedWalletRow.arriving, the funding form's "spend these coins again"
line, the member wallet screens' reach text (MemberWalletManager.delaysOf)
and the member certificate's proposal summary.
- Four of bennu's fixtures built a "guarded" wallet with no guardian leaf;
they now carry the ancestor's, as a verified statement would.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GroupWalletRollout.MEMBER_WALLETS_ON_MAINNET is on, lifted by decision
ahead of the signet run that was to lift it, as GUARDIANS_ON_MAINNET
was. A member's own spend on mainnet stays held to MAINNET_CAP, like
every group and guardian spend: memberCapFor no longer lifts with the
gate, since the run that would have justified lifting both has not
been made. The note says the gate is open, what it still leaves open,
and where the cap lifts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brings in a group funding a descendant's guarded wallet. Three conflicts:
- GroupSendScreen: funding replaced the address field and named the
funded wallet in the preview, inside the fields this branch moved
into SpendForm. SpendForm now takes a destination slot, which a
funding fills with its card, and the funded wallet's name; the
preview takes both that name and the member's ownWallet words.
- GroupWalletScreen: the guarded wallet row is funding's card with Fund
beside Spend, headed by this branch's #name / @name and carrying a
member's label.
- docs/README.md: both rows.
Funding reads guarded wallets through GuardianManager.guardedWalletsOf,
which this branch widened to members' wallets, so a group can now fund
a member's guarded wallet. Every place funding named a wallet with
`name ?: "a subgroup"` -- the title, the destination card, the preview,
the proposal summary, a signer's refusal -- uses displayName, so it
reads "Fund @alice's guarded wallet". A test funds a member's wallet
end to end through GroupSpend.fundingsOf.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The plan is marked built, with a table of the thirteen places the build
chose differently -- the key notice carrying the key rather than the
signed key state, the request travelling whole inside an inner event,
Put to the group living on the group's page, the wallet screen
following a member's spends while open, and the signet run still to
make before the mainnet gate lifts. The guardians and subgroups notes
point here, and the README says it has been built.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
guardian-delay.md is marked built, with a table of where the build chose
differently. group-wallet-guardians.md's "Why 12, and why for everybody"
becomes "Why a delay, and who sets it": each parent chooses its own leaf's
delay from 6 to 65,535 blocks, inherited ones keep theirs, and a long
delay is a backstop rather than an override. The policy note's picker
section and the README follow.
The three-emulator run is still to do by hand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A parent sets only its own leaf's delay; an inherited ancestor keeps the
delay it chose, all the way down a guarded subtree.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A parent picks its own leaf's delay when it makes a guarded subgroup, from
6 blocks to 65,535 (BIP68's block-based ceiling, about 15 months); an
inherited ancestor keeps the delay it chose. The delay rides the ceremony
proposal, is signed into the certificate's guardian tag and checked by
both quorums, and every screen that quotes "12" says each leaf's own
delay as blocks with an approximate duration. Eight phases, not built.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The member's device makes a random key, works out every ancestor leaf
and both descriptors, and only then asks the group. The group's quorum
signs a member wallet certificate of its own kind (30334) rather than a
birth certificate, since every reader of 30329 assumes a room. Ancestors
list the wallet as @name, and the group's page opens the member's profile.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
subgroup-wallet-policy.md is marked built, with a table of where the
build chose differently. group-wallet-guardians.md's "why it is still the
child's choice" becomes "whose choice it is": the parent decides, the
chain still does not enforce it, and an older certificate still leaves it
to the child. subgroups.md gains the wallet tag and the policy as the
second value the coordinator picks for other people.
The three-emulator run is still to do by hand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The coordinator picks guarded or unguarded on the subgroup picker, the
parent's quorum signs the choice into the birth certificate, and the
child's admins agree to it by joining and again when they sign the key
state. Every honest client then offers, signs, derives and watches only
the wallet the certificate allows. An unguarded certificate carries no
guardian keys, and a guarded group can only make guarded subgroups.
Not built: Phase 0's decisions and nine phases.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
FrostSigningEvents.DECLINE and MarmotDirectMessage.ENVELOPE_KIND are both
30330. Recognising protocol kinds first (ceremony-creation Phase 6) returned
before the direct-message branch, so every direct message in a
current-profile room was dropped unopened. A 30330 that holds a gift wrap is
an envelope; only one that does not is a decline.
ProtocolKindDispatchJvmTest now opens an envelope through
fromGroupEventResult, which no test did before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GroupWalletRollout.GUARDIANS_ON_MAINNET is on, lifted by decision ahead of
the signet run its condition named. The mixed-version risk stays open and
is said so in the gate and the guardians note; guardian spends on mainnet
are held to MAINNET_CAP like every other spend.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The group's detail screen loses its "Nostr profile", "Relays" and "Posts"
rows, and the six screens behind them go with them: the profile, its
editor, the relay lists, their editor, the posts and the composer. So do
the readers they stood on (GroupNostrProfile, GroupRelayList, GroupPost),
their ChatRepository methods, and 97 strings.
Two things go that were not named, because nothing else reaches them.
BroadcastGroupSignedEventScreen was only opened from those three screens,
so the app no longer sends a group-signed event to a relay, and
EventPublishTransport goes with it. "Propose event" accepted only a
profile, a post or a relay list, so it would refuse every paste;
ProposeGroupEventScreen and GroupEventProposal are removed, and with them
the detail screen's canEditNostrProfile gate. What is left of that block
is the wallet row, shown in Marmot rooms only.
The transcript keeps what it already holds. The three message types stay
as private legacy names in GROUP_IDENTITY_TYPES, as the curated-lists
types did, so rows already written stay system lines. A profile or relay
list arriving in a room now applies to nothing, so a signing session an
older build opened for one ends without a raw-JSON bubble. A kind:1 falls
through to "unsupported", which is what it was before posts existed.
ProposedEvent no longer describes the three kinds.
GroupSignedWorkRowsJvmTest now covers the wallet row and the absence of
the removed rows. The chronicle tests still use a profile, a relay list
and a post as their kinds; they compare GroupSignedEvent rows, which are
unchanged, so only ChronicleTestKinds' account of why is corrected.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Found on the emulators: refusing a Marmot room left the app bar reading
"Shared key for " and nothing. It reads "Shared key" until the room is
loaded. The plan now records what was checked on the three emulators that
found the bug, and what was not.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
shared-key-ceremony.md says where a ceremony starts and how one ends early;
subgroups.md points its Phase 10 here; member-chronicle.md says the chronicle
is built and carries nothing; the batch-signing, skipped-keys and Material
notes mark their document examples as removed; the plan records the ten
places the build chose differently.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Joining a subgroup's ceremony could open a second, whole-room ceremony in the
parent's Marmot room: the shared-key screen showed a ceremony, asked for
approvals and offered to start one from the same button slot. The plan
separates starting from showing, guards proposeRitual, and removes the
unreachable nip30303 document code whose kinds collide with the ceremony's,
keeping the member chronicle as an empty, quiet shell.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A guarded wallet's coins have a spender outside the room: a guardian can
take a coin the child's open spend wants, and the child can move one a
guardian's wants. Neither winner is a row in the loser's room, so the loser
sat signed and rebroadcast forever. The watcher now asks the chain, for a
signed spend Electrum has not heard of, which transaction took each coin;
once that one confirms, recordReplaced marks the spend replaced and settles
every other spend of the coin, as recordConfirmed does for the room's own.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The guardians note records what was built against the plan, what is left to
do by hand before the mainnet gate lifts, and the conflicting-spend gap. The
subgroups and wallet notes each change the one sentence the plan said they
would.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The parent signs a guardian key for the child into its birth certificate,
beside its own ancestors' keys derived one step further, and the child's
guarded wallet puts each in a 12-block leaf.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The audit and the JVM suite ran after every phase. The two-device regtest
walkthrough has not been done; the note lists each of its steps beside
the automated test that comes closest and what that test cannot show.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every place a spend is described -- the proposal list and the wallet's
spend rows through ProposedSpend.summarize, the approval screen through
isChange, and the signed and confirmed transcript lines through
amountSent -- already reported the payment rather than payment plus
change. Nothing needed fixing; ProposedSpendTest and the end-to-end
spend test now hold each of them to it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The amount field shows its unit, how many sats are confirmed and free to
spend, and on mainnet the cap; a problem with the amount -- not whole,
too large, more than the group has, too small, over the cap -- is said on
the field in its error state instead of under the form. The send-everything
row is one 48dp toggleable target, and turning it off again finds the
amount typed. The wording of every reason is one function shared by the
field and the preview.
GroupSendScreenJvmTest types into the form on regtest and mainnet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Marks the plan built, with a table of the nine places the build chose
differently, and points the group-chain-wallet note's gap-limit
paragraph and the docs index at it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SEND_AN_AMOUNT kept every network on send-all until a release had shipped
with no mislabelled change, which no release could do while the gate was
closed. It is removed: the send screen always offers the switch and the
amount, and on mainnet the existing 100,000 sat cap bounds what a change
bug could put at risk. Phase 3's end-to-end test is what stands in for
the release the gate was waiting for.
The view model test no longer branches on the constant, and a mainnet
test holds an amount to the cap.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two open spends with change both paid to the first unused change
address, since a spend waiting for signatures has not used it yet.
Change indices are now reserved like coins: reservedChangeIndices reads
the claims of every open spend, and the builder takes the lowest watched,
unused, unheld index. It chooses among generated addresses only, so
change never lands outside the window; CHANGE_LOOKAHEAD rises to 10 and
past it a spend with change returns TooManyOpen.
The change branch not having answered is FindingChange rather than
NothingToSpend, and change too small to send back is reported as
Built.absorbed and said in the preview.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The gated amount field read its text by dropping every non-digit, so
21000.5 became 210,005 sats and 0.0005 became 5. parseSats accepts
thousands grouping and refuses decimals, grouping that groups nothing,
non-ASCII digits and more sats than exist; the preview reports which.
The field shows plain digits grouped and keeps what was typed.
Also plans the rest in docs/group-send-an-amount.md: send-an-amount
opens on every network at once, held on mainnet by the existing cap.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Modelled on Phoenix's swap-in addresses screen: every paid receive
address, the current one, and the two after it, read off the receive
mini-wallet's state with nothing stored and nothing new subscribed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
And its registry, GroupChainWallets, is GroupOnchainWallets, with the
files, the test and the plan's references following.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GroupWalletRollout holds the three gates, each with what lifts it. The
flag is on, its condition met. Send-an-amount is off, so the form sends
everything confirmed to one address and says so. A spend proposed from
this device on mainnet may send at most 100,000 sats; test networks
have no cap. The gates are on the composer and nothing else: a device
behind them still accepts, shows and signs what another member
proposes, through every check GroupSpend makes.
The plan is marked built, with its table of where the build chose
differently. Full JVM suite: 1474 tests, none failing; the Android
target compiles; every M3 budget is met.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every spend the room proposed, newest first, for every member whether
or not they have a node: what it sends and where it has got to --
waiting for signatures, abandoned, signed, sent (reported back by
Electrum, not merely broadcast), confirmed, or replaced by a later
spend. A signed spend Electrum has not reported can be sent again; a
signed, unconfirmed one can be bumped to twice its feerate, a new
session over the same coins that whichever confirms settles. Each row
opens its signing session.
Full JVM suite: 1473 tests, none failing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GroupSendRoute and GroupSendScreen, behind a Send button on the wallet
screen for a member who holds a share and has a node to see the coins
with. Address, amount or send everything, and a feerate prefilled from
Electrum's six-block estimate. The preview is GroupSpendBuilder over the
same inputs the proposal is built from -- what leaves, the fee, what
comes back as change, or why it cannot be built -- and proposing moves
onto the signing session, replacing the form.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Kind 30332 opens a session over a PSBT, one item per input, each its own
FROST instance over that input's BIP341 sighash with its own nonce seed
and its own tweak cache. proposeSpend and acceptSpendProposal run the
same GroupSpend checks: every input derives under the account the
room's verified statement gives -- looked up under the ceremony's own
key -- default sighash only, no script-path fields, replaceable, change
only where its claim derives, and a fee under a ceiling. A second
proposal under one id is ignored. completeSpend checks each signature
against its coin's output key and the whole transaction with
correctlySpends before writing it.
An old build does not route 30332, so it never signs a transaction it
cannot show. Resend repeats the PSBT byte for byte. Two-device test:
a quorum spends two coins and both devices hold a transaction that
correctlySpends; each check refuses a hand-built proposal outright.
Full JVM suite: 1458 tests, none failing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
FrostSigningSession.subject, NOT NULL with a default of EVENTS, says
what a session's items are messages of; FrostSigningItem.inputIndex
names the input a transaction item signs, with message and payload as
getters over the existing columns. GroupTransaction holds a spend from
proposal on: the PSBT, written once with IGNORE, the txid fixed before
signing, and the signed transaction, broadcast, seen and confirmed.
AutoMigration(25, 26) is three ADD COLUMNs and a CREATE TABLE -- no
rebuild of either signing table, so no cascade fires.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GroupWalletRoute and GroupWalletScreen, reached from a Group wallet row
on the group's screen. A group that has not opened a wallet gets an
empty state with, for a member who can sign, what opening one asks of
everybody -- including that shares are unencrypted at rest -- and the
button that proposes the statement. An open wallet shows the balance,
confirmed and on the way said apart, the next receive address as a
BIP21 QR code, selectable text and a copy button, the coins, and an
export of the two watch-only descriptors with what they give away.
The wallets are watched on the identity's own Electrum client and made
per PhoenixBusiness in BennuNavHost; a profile with no node sees the
wallet and is told it cannot watch the chain.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GroupChainWallet is Phoenix's FinalWallet for a group: two
ElectrumMiniWallets, one per BIP44 branch since a mini-wallet holds one
generator, gap-limited at 10 for receive and 3 for change, on the shared
client. GroupWalletState gives the coins with the derivation that spends
each, the confirmed and pending balances, the next receive address and
change index, and derivationOf -- bounded by the addresses the wallet
has generated, which is how far a check of a claim will look.
Tested against a fake IElectrumClient: a payment moves the receive
address on, change is counted and never offered, and two groups on one
client see only their own coins.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Kind 30331, inner only: the account xpub, the network, the account
number and both descriptors. A reader believes one only when the room
signed it and the room's threshold key rederives every value it
carries -- GroupWallet.of, used by the repository and by the transcript
arm alike. The broadcast screen refuses the kind twice, at the loader
and in sendToRelays, because the xpub on a relay is the whole balance.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GroupWalletDerivation: a public account step -- a tagged HMAC over the
threshold key, the network and an account number -- gives a depth-0 xpub,
and below it BIP86's own unhardened tail, /<change>/<index>, with the
key-path tweak on top. Each address is computed twice, as points and as
tweaks on a FROST cache, and refused if the two disagree.
Tests: BIP86's published vectors through the cache prove the tail; a
pinned vector freezes the account step; a sweep covers both parities.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Modelled on Phoenix's final wallet, phased. The path is BIP86's
unhardened tail under a public account step over the ChillDKG key, so
the group can export a standard tr() descriptor; each input is a batch
item with its own tweak cache, and a member checks seven things before
signing a spend. Shares are used as they are at rest, for now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Move the packages to com.machankura.bennu.{compose,android,desktop}, the
Android applicationId to com.machankura.bennu.android and the iOS bundle
id to com.machankura.bennu.ios. The root project, launcher label, iOS
product, desktop window, user-facing strings, client tags and the
Curare* classes all say Bennu now.
Left alone on purpose: the mantra/ derivation tags, the `curated` desktop
data directory, the mantra.press relay and the curare.to references.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>