`cat > "$WORK_DIR/Dockerfile.rootfs" <<DOCKERFILE` was unquoted, so the build
shell performed command substitution on the Dockerfile body. Any backtick in a
Dockerfile COMMENT was executed on the build host and its output spliced into
the generated file. Six comments did this. One of them ran
`systemctl start archipelago-fips.service` against the build machine on every
ISO build; the others were harmless only by accident of being command-not-found.
Fixes the class, not the six instances. The delimiter is now quoted, so the
body is emitted verbatim and a future backticked comment is inert. Verified the
boundary by line range first: the other backticked comments in this file
(:264, :809, :1188, :1289, :1506, :1605, :3597, :3651) are ordinary shell
comments outside any unquoted heredoc and were never at risk — they are
untouched.
The body needs exactly four build-time values and they are all package names
(LINUX_IMAGE_PKG, GRUB_EFI_PKG, GRUB_EFI_SIGNED_PKG, GRUB_PC_PKG), on four
consecutive lines. So quoting was practical: the heredoc is split into
DOCKERFILE_HEAD and DOCKERFILE_TAIL, both quoted, with a single explicit printf
interpolating those four names between them. Escapes that existed only because
the heredoc was unquoted are undone in the same pass: six trailing `\\` become
`\` (Docker line continuations) and four `\$` become `$` (RUN arguments reach
the shell verbatim — Docker does not substitute variables in RUN).
Verified by rendering the generated Dockerfile before and after with the same
inputs and diffing them normalised (continuations joined, whitespace
collapsed). Both are 190 normalised lines and the ONLY differences are the six
comments regaining their text — every instruction is byte-identical. Before:
"# the archipelago backend calls" / after: "# the archipelago backend calls
`systemctl start archipelago-fips.service`".
Test: case 7 asserts every heredoc writing Dockerfile.rootfs has a quoted
delimiter, and when one is not, reports which body lines would execute. The
assertion is on the delimiter, not on backticks — with quoting a backticked
comment is legal and six of them are back in the body on purpose, so flagging
backticks would flag a non-bug and fail on the very comments this restored.
This bug is invisible to `bash -n`; an instance of it introduced earlier in
this plan hung a syntactically-clean build for two minutes before being caught.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Unify rather than delete. The defect in F-03 was never "a second attempt to
create a key exists" — it was that failure was silent and the completion marker
lied about it. A second attempt is only dangerous when it is an unaudited second
PRODUCER carrying its own idea of success, its own absent retry policy and its
own absent failure record.
Single producer. gen_tls() is now the only code in the ISO build that creates
/etc/archipelago/ssl/archipelago.{key,crt}; gen_ssh() the only code that creates
/etc/ssh/ssh_host_*. Two secondary producers are gone:
- the Dockerfile's `openssl req` layer, which baked a keypair the strip layer
deleted moments later in the same build;
- the installer's "ensure SSL cert exists for nginx HTTPS" block, which before
the strip almost never fired and after it would have fired on every install.
Proof is mechanical, not a claim: every executable `openssl req` / `ssh-keygen
-A` invocation in the builder now lives inside the generator heredoc, and the
test suite fails if one appears outside it.
Build-time assertion. The one realistic total failure is a missing generator
binary, which is deterministic — no retry or reboot fixes it. A rootfs RUN layer
now fails the build if openssl or ssh-keygen is missing or non-executable.
openssl and openssh-server are both already in the package list (and
openssh-server hard-depends openssh-client, which ships ssh-keygen), so today
this is cheap insurance; it earns its place the first time someone edits that
list.
Self-heal, never dead-end. Fail-closed governs SERVING; retry governs
RECOVERING, and they are different things. Adds
archipelago-first-boot-secrets.timer (OnBootSec=5min, OnUnitActiveSec=15min),
installed and enabled with a hand-written symlink fallback because chroot
systemctl enable can fail silently. The service's own ConditionPathExists=!
makes every trigger a no-op once the marker exists, so a healthy node pays
nothing. On success the script now restarts consumers that are in `failed` —
try-reload-or-restart is a no-op on a failed unit, so without this a recovered
node would have valid keys on disk and nginx still down.
Never serve a bogus key. gen_tls parses both halves back with `openssl pkey`
and `openssl x509` before the swap, so a truncated or half-written artefact is
never what nginx reads.
Tests: 6 cases, each with an isolated negative control (transcripts in SUMMARY).
- case 4, TLS fails every attempt on a stripped root -> no key from any source.
Control: reintroduce a fallback key creation -> only case 4 red.
- case 5, self-heal: a failed run then a later successful run -> key present,
marker set, failed units restarted. Control: dead-end on a node that already
failed -> only case 5 red.
- case 6, single-producer invariant. Control: reintroduce the installer block
-> only case 6 red, naming the line.
Residual risk, stated plainly: a machine where generation can never succeed
still ends up with no SSH and no TLS. Build-time assertion removes the
deterministic cause, retry plus timer removes the transient ones, so what
remains is genuinely broken hardware — and it says so on the console and in
/var/lib/archipelago/first-boot-secrets.failed rather than quietly serving a
key nobody audited.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A guide a third-party security auditor can use to verify Phase 10's claims
without trusting our test harness — and that we use ourselves.
Every claim carries four parts, all required: the claim stated falsifiably;
how to REPRODUCE THE DEFECT on the parent commit; how to verify the fix; and a
negative control that must go red on exactly that defect and nothing else. A
test passing on both fixed and unfixed code proves nothing, and reproduce-first
is the step most often omitted in security theatre.
Prefers external checks (curl from another host, tar listing, cross-node file
comparison) over our own tests wherever a claim can be checked from outside.
Tiered by hardware needed: Tier 0 any checkout, Tier 1 running node, Tier 2 ISO
build host, Tier 3 two physical nodes, Tier 4 pre-release gate. Status marked
per claim — verifiable now, pending a plan, or hardware-gated — so an unmarked
absence is never read as a pass.
States what is explicitly NOT claimed (Lightning custody is not air-gappable;
no claim against a compromised kernel CSPRNG or supply chain; KEY-05 is
structural not exploitable), the known-accepted risks with where each was
decided, and carries the C-6 warning that probing with seed.status reports the
surface closed while the real door stands open.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Task 3 of 10-03 is a blocking checkpoint: proving the shipped rootfs tar is
identity-free needs a real ISO build host with podman/docker and disk for a
full rootfs rebuild. This commits the prepared evidence document with the exact
command sequence, marked UNVERIFIED, rather than claiming the check passed.
The document states the inverted expectation explicitly. The audit's C-4 entry
expected SSH host keys and the TLS key to be PRESENT — that described the
broken state it was measuring. After the strip layer those must be ABSENT, so
the audit's stated expectation is now the failure condition. A future reader
comparing the two would otherwise conclude the check regressed.
Also records two things the operator would otherwise get wrong:
- RECIPE_HASH must be read from the stamp file, not computed from the repo
file. build-debian-iso.sh rewrites the builder's relative paths into a temp
copy before exec, and the hash covers "$0"; the hashed region has 35 such
rewritten expressions plus an absolutised SCRIPT_DIR, so the value is
specific to the build host and checkout path.
- C-4 is a build-host check only. Two-node key divergence is C-3 and stays
separately UNVERIFIED; the note explains why SSH host keys are the sharper
signal there than TLS, given the installer's per-install TLS fallback.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The rootfs is a container image exported to a tar and extracted verbatim onto
every disk flashed from the ISO, and the ISO is published. It baked two things
nobody asked for: Debian's openssh-server postinst generates /etc/ssh/ssh_host_*
during the container build, and the `openssl req` layer writes the TLS keypair.
Both were therefore identical on every node and known to every downloader.
Add a final RUN layer to Dockerfile.rootfs that removes /etc/ssh/ssh_host_*,
removes the archipelago TLS keypair (keeping the ssl directory so the first-boot
staging swap has somewhere to land), truncates /etc/machine-id to systemd's
documented "regenerate on next boot" state, and drops a non-shared
/var/lib/dbus/machine-id if one exists as a real file rather than a symlink.
It also writes /opt/archipelago/rootfs-identity-stripped so a node can answer
after the fact whether its rootfs came from a stripped build; no timestamp,
so the RECIPE_HASH cache stays reproducible.
This is what makes 10-03's fail-closed regeneration structural instead of
procedural: with the material gone, a regeneration failure degrades to
"no key, service refuses to start" rather than "fleet-shared key, silently".
The `openssl req` layer is deliberately left in place — it keeps proving
openssl is present and keeps the SAN template next to its consumer; the strip
layer is what makes the output non-shared.
Two comment corrections that follow from the strip:
- The installer's TLS block is no longer a rarely-taken safety net; it now
fires on every install. It is per-install and never image-wide, so it does
not reopen F-03, but it does mean a first-boot failure still leaves the web
UI with a cert while SSH has nothing. Comment updated to say so.
- The first-boot script header overstated the fail-closed cost for TLS for the
same reason; corrected to claim certainty only for SSH.
This edit is inside the RECIPE_HASH region, so the next build is forced to
rebuild the rootfs tar — required for the C-4 evidence to mean anything.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The first-boot per-device secret regeneration was fail-open: both branches
logged a warning and continued, and `touch "$MARKER"` ran unconditionally
outside both `if` blocks. Combined with the unit's ConditionPathExists=! and
the script's own marker fast-path, one transient failure left that node on the
image-wide shared SSH host key and TLS private key permanently and silently —
and the ISO is a published artefact, so every downloader holds those keys.
- Retry each generator 3 times with backoff (D-05), so a transient first-boot
condition recovers inside the same boot instead of being terminal.
- Write the completion marker ONLY when both TLS and SSH succeeded, so a
failed boot leaves the unit eligible to run again on the next boot.
- On terminal failure: durable record at
/var/lib/archipelago/first-boot-secrets.failed naming which generator
failed, plus console + logger + stderr, and exit 1 so the unit lands in
`failed` rather than `active`. The record is cleared on a later success.
- Add FIRST_BOOT_SECRETS_ROOT / FIRST_BOOT_SECRETS_BACKOFF seams. Unset in
production the behaviour is byte-identical; set, they let the fail-closed
property be asserted rather than claimed.
- Order the unit After=systemd-random-seed.service (no-op today, correct if a
seed file is ever baked).
- State the operational trade in the script header: after the rootfs strip, a
terminal failure means no SSH and no TLS and needs the physical console.
That was chosen deliberately over running on fleet-shared keys.
tests/first-boot-secrets/run-tests.sh extracts the shipped heredoc body from
the builder and drives it against a temp root with stubbed generators: both
succeed, openssl fails every attempt, ssh-keygen fails twice then succeeds.
Moving the marker touch back outside the success branch makes case 2 fail with
MARKER-SET-ON-FAILURE, which is the regression this pins.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
98 operational files still carry 146.59.87.168 — the domain was only adopted
for the git remote, not for container registry references. Bulk is app
manifests' image: lines, plus .gitmodules, both CI workflows, the signed
catalog.json, and two Android companion files with compiled constants. The 117
hits in .planning/ are historical records and stay.
Not a find-and-replace: the domain serves Gitea over HTTPS:443 while images are
pulled from :3000 over plain HTTP, and podman treats host:3000 and domain as
different registries — so every node re-pulls under the new name and any node
that can't resolve or trust the new host fails to pull. It also invalidates the
signed catalog (needs a re-sign ceremony) and the APK ships compiled constants.
Rollout order: registry serving on the domain → manifests → catalog re-sign →
APK rebuild. Steps 2-4 are actively breaking until step 1 holds.
Analysis from the concurrent agent's session before it ended; recorded so it is
not lost.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Companion build 0.5.27 (versionCode 47) shims navigator.clipboard natively, so
in-app copy/paste is fixed with zero web changes — but the contract must not be
clobbered (no unconditional re-define, no Object.freeze).
Still open web-side: main.ts's fake readText() makes SendBitcoinModal's Paste
button render and silently no-op in plain-HTTP browsers; 30 writeText call
sites across three inconsistent patterns, ~10 of which toast 'Copied!'
regardless of success; scanner prewarm/torch/constraints/no-reinit.
Also records three factual corrections to docs/qr-scanner-snappiness-handover.md
(ZXing not ML Kit; FORMAT_QR_CODE + KEEP_ONLY_LATEST already in place; do NOT
drop to 720p — 1080p is a deliberate 0.5.22 fix for dense bolt11 QRs).
Routed at Phase 11: the signed-PSBT paste affordance and the scanner items are
the same surface as WALLET-05.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Clicking "Open a channel" or "Setup Guide" navigated correctly but left the
wallet's send/receive modal floating over the destination. The Lightning modal
itself did close — the parent did not. Tab views are KeepAlive'd, so
navigating deactivates the owner rather than unmounting it, and its Teleported
modal keeps rendering.
BaseModal now emits close on any route change while shown, fixing the class in
one place rather than per button. Every modal here is a transient dialog; none
should survive navigation. Two tests pin it, including that a hidden modal
stays quiet.
Also fixes a test-only regression from fa26c5fc: useLightningRequired()
resolved the Pinia store at composable-call time, so merely having the gate in
SendBitcoinModal made it unmountable without an active Pinia (PaidTick mounts
it bare). The store is now resolved lazily inside the function that needs it —
a gate should never be what breaks a component's ability to mount. That one
shipped because I verified fa26c5fc with targeted tests and a build but had
not re-run the full suite since 5718179e.
Verified: full suite 103 files / 827 tests green; npm run build clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A running LND with zero channels happily mints an invoice — it is simply
unpayable, because nobody has a route in. So the state-only gate let receive
through and handed the user a useless invoice, and let send walk to confirm.
Neither errored, so the funding modal (wired to failures) never fired.
requireLightningReady(direction) now asks lnd.listchannels and checks the
liquidity that actually matters for the attempt: total_inbound to receive,
total_outbound to send. It fails OPEN on an RPC error — a transient blip
should not block a working wallet.
The no-funds mode says plainly that a channel is needed, in the direction's
own terms (inbound vs outbound), and offers both routes: "Open a channel"
straight to the channels screen where the Zeus/Olympus flow is already
prefilled, and "Setup Guide" to the run-lightning-node walkthrough for someone
who wants the whole path explained. Buttons wrap rather than squeeze on
narrow screens.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Icons: each node choice now shows its app icon (lnd.png; Core Lightning's is
vendored from the Umbrel gallery as core-lightning.svg). Vendored rather than
hotlinked on purpose — these nodes run offline/airgapped, and a remote image
would both break there and leak a request to a third-party host on every
render. A missing asset falls back to a neutral bolt glyph so a row can never
render a broken-image box.
Mobile: the choice row keeps icon + name + blurb together and drops the action
to its own full-width line under 26rem, instead of squeezing the description
into a two-word column next to a button.
Funding mode: a node that is running but has no funds / no inbound liquidity
is neither "install one" nor "start it", so the same modal gains a third mode
that explains it and routes to the run-lightning-node goal, where funding and
channel-opening already live — reusing that flow rather than duplicating it.
It fires where the user actually meets the problem: on a failed attempt.
handleLightningFailure() maps a running node's send/receive failure onto the
funding modal, matched on message text because LND surfaces "no route", "no
channels" and "insufficient balance" as plain strings with no distinct code —
and all three mean the same thing to a user: fund me.
Verified: 5 gate tests; npm run build clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Three defects from testing the previous commit on archi-dev-box:
1. The gate keyed on `id in packages`, which is not "installed and usable" —
package-data carries an entry for a Lightning app that is known but not
running. On a box with no lnd container at all the gate passed and the raw
error came through as "Operation failed. Check server logs for details."
Now keyed on PackageState.Running.
2. Because installed-but-stopped is a real and different situation, the modal
has two modes: absent offers the install choices, stopped says the node
isn't running and offers "Open My Apps". Neither dead-ends in an error.
3. Lightning SEND let you walk all the way to confirm-send with no node. The
gate now runs in review(), before the confirm step — failing at submit
after a review screen is the defect, not a smaller version of it.
Also adds CopyButton, the start of one consistent copy affordance: icon +
label, an emerald tick held 1.6s, a fixed box so the width never jumps, and a
document.execCommand fallback so copy still works over plain http on a LAN IP
(navigator.clipboard rejects on insecure origins, which is how a lot of nodes
are reached). Converted the wallet's own copies — the lightning invoice the
user reported, plus the on-chain/Ark addresses and the payment hash/txid.
20 of 25 copy sites across 15 other files still use ad-hoc markup; converting
them is mechanical but was not attempted here rather than half-done.
Verified: 5 gate tests; npm run build clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The F-10a scope correction committed hours earlier asserted semantics its
evidence did not support. The KEY-05 planner caught it against the code:
- mesh/x3dh.rs:100/:114 are u32 prekey IDENTIFIERS (spk_id, otk_id), not key
agreement material. The X25519 secrets come from
crypto::generate_x25519_ephemeral() at :99/:113 and were never in scope.
- session.rs's 16 raw matches read as 16 production token sites; #[cfg(test)]
begins at :470, so it is 4 production + 12 test.
- wallet/bdhke.rs is 2 production of 4 (#[cfg(test)] at :143) — and those two
ARE genuine key material: generate_secret() :133 and
random_blinding_factor() :139.
The Medium rating still holds, on narrower grounds: bdhke's two production
sites plus storage_crypto.rs:39's AEAD nonce. It no longer rests on x3dh.
Struck rather than silently rewritten. F-10 was corrected on the grounds that
understatement misleads the next reader; overstatement does the same, and this
table managed both within a day.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Adds 10-06-PLAN.md covering KEY-05 (F-10a / R-16). 10-01..10-05 untouched.
Six tasks, sequenced so CI stays green at every intermediate commit:
classify all 43 call sites with file:line evidence; a tracer that wires the
sealed KeyGenRng allowlist, the degenerate-entropy predicate and the CSPRNG
readiness ledger end-to-end through the mnemonic seam; two migration tasks;
a blocking human checkpoint for cargo-deny scope and legitimacy; then the
gates are enabled last and observed failing a real build.
The clippy ban is a compile failure under the existing -D warnings CI step,
so core/clippy.toml is deliberately not committed until every site --
including test code, since --all-targets counts it -- has migrated.
Wave 2: shares seed.rs with 10-05 and api/rpc/auth.rs with 10-01.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Creating a Lightning invoice with no Lightning implementation installed failed
at the RPC layer — lnd.createinvoice returned connection-refused and the
Receive screen rendered it as a red error. That reads as the wallet being
broken when the node simply has no Lightning node installed yet.
useLightningRequired() gates the three invoice paths (wallet Receive, the Web5
send/receive sheet, and the app launcher's paywall — both arms there, since
paying an invoice needs a node as much as minting one). With none installed it
raises a modal offering to install one and the caller bails without surfacing
an error at all.
The modal lists the choice rather than assuming LND: LND installs today, Core
Lightning is listed greyed as "Coming soon" so the platform doesn't read as
LND-only. When CLN ships it is two lines — flip `available` and add the id to
LIGHTNING_NODE_APP_IDS.
Detection is install state, NOT reachability, deliberately: an installed node
that is merely stopped or still starting is a different problem ("start it")
and must not be answered with "install a Lightning node".
Also fixes the credentials modal, which painted its own rgba(8,10,18,.98)
navy card instead of the house glass-card — it read as blue against every
other modal. It existed twice (Apps.vue and apps/AppIconGrid.vue); both now
use BaseModal, so they also inherit Esc/focus handling, body scroll lock and
the standard pinned-header/footer scroll contract they were missing. Dead
panel CSS removed from both.
Verified: 4 new tests; full suite 103 files / 826 tests green; npm run build
clean with the new strings present in the built bundle.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The checkpoint on archi-dev-box proved rotation alone doesn't close FED-07:
the credential file went unique while the RUNNING container kept serving the
compromised one, because the Quadlet path rewrites a unit without restarting
it and fedimint-gateway is classified restart-sensitive, so drift was detected
and deliberately ignored on every tick.
Rotation now records the app id, and the drift check consumes that flag to
recreate even a restart-sensitive app, with a WARN naming the reason. This
mirrors the published-port carve-out a few lines above, which already makes
the same trade for the same reason: a container that is already broken (there)
or already compromised (here) is not protected by leaving it running.
Restart-sensitivity protects working services. A gateway answering to a
credential published in this repository is not working, it is compromised, and
gateway admin can drain Lightning liquidity — indefinite exposure loses to a
few seconds of restart. Rotating-but-only-alerting was rejected: the
monitoring system fires on metric thresholds only, so it would have needed new
event-alert plumbing to deliver something strictly weaker.
Re-verified on the same node, same scenario: rotation at 06:39:23, recreate at
06:39:27, PID 3923125 -> 148426, running credential now matches the file,
container healthy with the same name and ports, gatewayd.db intact at 18 files
with IDENTITY present, 32 containers untouched, no repeat rotation.
3 new tests. Also lands the missing 01-19 and 01-20 SUMMARYs: both had code
committed 2026-07-31 but no summary and no roadmap tick, so they read as
unstarted. Phase 1 is 11/20. FED-09 carries 15h of Tor uptime and 0
permission-fixes across 542 doctor runs on archi-dev-box.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The audit recorded F-10 as two call sites in container/secrets.rs. The real
defaulted-RNG surface is 41 sites across 15 files: session.rs (16),
pine_ha.rs (6), wallet/bdhke.rs (4 — ecash key material), mesh/x3dh.rs (2 —
key-agreement material), storage_crypto.rs (1 — AEAD nonce), +10 more.
Nothing is broken today: rand::random()/thread_rng() are ChaCha12 seeded from
getrandom(2). What changes is blast radius — F-10's Low rating rested on
'per-app credentials rather than the master key hierarchy', which does not
survive the true scope. Re-rated Medium as F-10a.
Records why the original audit missed it: F-10 was reached by tracing the
manifest-secrets path, and no step enumerated defaulted-RNG use across the
crate independently of the traced paths.
F-10's original text is left unedited so the correction is auditable rather
than retroactive. R-13 superseded by R-16; tracker item replaced.
Adds KEY-05 to Phase 10: sealed allowlist trait at key-gen seams, clippy
disallowed-methods ban (compile-time, CI-enforced), cargo-deny on duplicate
rand majors, degenerate-entropy runtime check, persisted CSPRNG-readiness
verdict. Also retires the false 'impl CryptoRng for CountingRng' at
seed.rs:656.
Records the user's execution gate: Phase 10 does not start until the
concurrent Phase 1 agent is finished and their changes are synced. KEY-05 is
unplanned — the existing 5 plans predate it and a 6th is required.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Handoff for the reactive demo-day session that followed 09-06/09-07
(both already complete). Covers: security audit (6 IDOR fixes across
the botfight repo), Cashu payout claim UI, existing-bot AI-config UI,
botfights 1.2.11 built+deployed to both demo nodes, catalog
signed+published.
Also: discovered and fixed 4 botfight-repo commits that were local-only
and never pushed to origin — pushed as part of this handoff step
(botfight @ d00e792..10d4209 -> origin/main).
Co-Authored-By: Claude <noreply@anthropic.com>
The executor was instructed to leave docs artifacts to the orchestrator; this
commits them: the research that drove the audit, the task summary, and the
archi-dev-box test-node todo raised during the same session.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Ran Task 2's blocking checkpoint on a real node. The rotation works and was
proven end to end: it fired ~15s after restart, wrote a fresh unique
credential (0600, service-owned), logged exactly one line naming the .pw path
with no value in it, left every other secret and the gateway's data untouched,
and kept the container's name and ports.
But the assumption the plan rests on is WRONG, and the checkpoint is what
caught it. 25 minutes after rotating, /proc/<pid>/environ showed the running
gatewayd still using the PRE-ROTATION credential while the file and podman
secret held the new one. The orchestrator explains itself in its own logs:
Quadlet unit drift-synced — file rewritten, .service NOT restarted
(operator restart picks up new config)
container drift detected during boot reconcile;
leaving running restart-sensitive app untouched
Two deliberate guards: the Quadlet path never restarts a unit it rewrites, and
fedimint-gateway is classified restart-sensitive so drift is detected on every
tick and then ignored — logged at 15:51, 15:53, 15:54, 15:56 and counting.
So on a real affected node the credential file becomes unique while the
gateway keeps answering to the compromised one until an unrelated reboot, and
the operator reading the .pw gets a password the gateway rejects — T-01-77
inverted. FED-07 is NOT closed and this plan alone cannot close it.
Not hand-rolled around, per the plan's own instruction. The fix needs a design
decision: whether a compromised credential is the case that should override
restart-sensitivity, or whether rotation must raise an operator-facing
"restart required" alert instead of logging into the void.
Incidentally disproved: restarting archipelago does NOT kill containers here
(29/29 then 31/31 survived; "Adopted 31 existing container(s)"). The service
is system.slice/KillMode=control-group while containers live in
user-1000.slice/…/libpod-*. The CLAUDE.md SIGKILL rule predates Quadlet.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
FED-07 was marked Complete when 01-11 landed, which was premature: the
requirement text explicitly includes "existing installs with the default
password get a migration path", and that migration has never been exercised on
a node. Corrected to code-complete/verification-pending.
Checkpoint step 1 was run read-only on archi-dev-box: the node is CLEAN (hash
present, 600, service-owned, not the shipped default) and has NO gateway
container — the app is installed but nothing runs and its data dir is empty.
So rotation cannot fire naturally here, and the steps that matter most (data
survives the recreate, new credential authenticates, old one rejected) have
nothing to exercise without installing and seeding first.
Deferred deliberately rather than run unattended: 30 containers are up with
4-8 days uptime (IndeeHub, Immich, BTCPay, netbird, strfry, …), the
archipelago system service is active, and restarting it SIGKILLs containers
until Quadlet is the default.
The todo carries the full context plus two adjacent findings: fedimint-gateway
is missing from handle_package_credentials (so a rotated password has no UI
retrieval path), and photoprism ships a fixed admin password in its manifest.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
01-11 stopped new installs from ever taking a shipped credential, but did
nothing for the nodes that already did — those gateways still answer to a
password published in this repository.
rotate_compromised_gateway_credential() detects an EXACT match against the
denylist and replaces the pair; absent, unique, or merely unrecognised values
are left alone and return false. That distinction is the point: an operator
who deliberately set their own credential also has an "unrecognised" one, and
rotating it would be the same class of harm as leaving the default in place.
It hangs off resolve_dynamic_env beside ensure_generated_secrets, gated on the
gateway's app id, so an affected node heals on its next reconcile tick. There
is deliberately no teardown here: the new hash changes the resolved secret env,
which changes secret_env_hash, which the drift check reads as a container-label
mismatch — so the platform's own recreate path rebuilds the gateway around its
unchanged data directory, ports, volumes and name.
Rotation is self-terminating (the value written is not on the denylist, so the
next tick is a no-op) and errors propagate rather than being swallowed, because
the atomic write leaves the previous credential intact on failure.
Bcrypt generation was factored out of ensure_one into write_bcrypt_pair, which
both generation and rotation call — 01-11's SUMMARY claimed such a helper
existed but the arm was still inline, and rotation cannot reuse
ensure_gateway_credential because its idempotent fast path returns early
exactly when the file is present, which is the case rotation acts on.
Also fixes cargo fmt drift left by 42652547 in install.rs.
Verified: 6 new tests, secrets suite 16/16; full suite 1008 passed with one
known wall-clock flake (green 4/4 in isolation). NOT verified on a node —
Task 2's blocking checkpoint has not been run, so FED-07 stays open.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The two-scan dance (node shows unsigned PSBT as animated QR, signer signs,
node scans the signed PSBT back, finalize + broadcast) is ~80% plumbed and 0%
usable. Verified gaps:
- No UI: lnd.create-psbt / lnd.finalize-psbt and their rpc-client.ts:417
wrappers are called by nothing but unit tests.
- No animated-QR encoder: qrcode/qrloop are deps and the inbound path
(useAnimatedQRDecoder + WalletScanModal) works, but nothing encodes a PSBT.
- Wrong format for real signers: qrloop is Ledger's; Passport/SeedSigner speak
BC-UR (ur:crypto-psbt), Coldcard Q speaks BBQr. BC-UR is the priority given
the existing Passport-Prime-compatible SeedQR work.
Gated on 10-05: create-psbt funds from LND's own wallet, so until LND is
watch-only against the external signer the offline device signs inputs whose
keys the node already holds.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
First-run wallet-type chooser, seed handling reusing the shipped SeedQR +
seed-words components, and evidence-based umbrelOS LND UI parity.
Gated on Phase 10's 10-05: the set of wallet types WALLET-01 can offer is a
direct consequence of the watch-only verdict that plan produces, and 10-05 also
deletes the dead Core wallet path so this phase never represents it in the UI.
Records the already-shipped inventory (channels panel, send/receive/scan/
settings modals, SeedRevealPanel, LndSeedBackupPrompt, utils/seedqr.ts) so the
parity matrix closes real gaps instead of rebuilding existing surfaces.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
D-03 named NodeIdentity::key_exists as one of the two gate signals. Server::new
(server.rs:63-72) calls load_or_create on both branches, and load_or_create
(identity.rs:48-51) generates and writes a random temporary node key when none
exists — so key_exists is true on any node that has booted once, onboarded or
not. A gate keyed on it would refuse seed.generate on a fresh node and brick
onboarding fleet-wide.
The flaw came from the audit's own suggested remediation (§214-221) and was
repeated in the planning brief; the planner caught it against the code.
D-03's intent (two signals, OR-ed, fail safe on drift) is unchanged. Corrected
signal set: is_setup() / is_onboarding_complete() / seed_exists(), pinned by a
test rather than a comment.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Closes the three exploitable findings from the 2026-07-31 entropy/seed audit.
Wave 1 (parallel):
- 10-01 KEY-01: shared onboarding gate refuses seed.generate/seed.restore/
seed.save-encrypted/backup.restore-identity/auth.setup on a provisioned node,
plus an auth.onboardingComplete guard and retry-budget-derived rate limits.
Independently shippable (D-11): no depends_on, no shared files.
- 10-03 KEY-02: first-boot secret regeneration retries with backoff then fails
closed; rootfs tar ships identity-free so failure degrades to "no key".
- 10-05 KEY-03: delete the uncalled bitcoin.init-wallet-from-seed xprv-import
path (D-07b); make LND's PSBT round trip first-class with a key-origin report.
Wave 2:
- 10-02 (deps 10-01) KEY-01/KEY-04: on-node C-6 exposure measurement, live
refusal proof, fresh-node onboarding non-regression.
- 10-04 (deps 10-03) KEY-02/KEY-04: fleet detection of image-baked host secrets,
guarded rotation behind a D-06 decision checkpoint, C-3 two-node verification.
Planning-time scoping correction recorded in 10-01: D-03 names
NodeIdentity::key_exists as the on-disk "onboarded" signal, but server.rs:63-71
writes a temporary node_key on every boot, so that signal is true on fresh nodes
and would brick first-boot onboarding. D-03's dual-signal intent is preserved
with is_setup / is_onboarding_complete / seed_exists.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Core's wallet is outdated and used by nothing, so bitcoin.init-wallet-from-seed
is deleted outright rather than migrated: uncalled, authenticated and
password-gated, never ran on archi-dev-box, and its only job is deriving and
stringifying the master BIP-84 xprv.
D-07's parity-proof migration and its one-way checkpoint are withdrawn — there
is no wallet to migrate. A small discovery check folds into KEY-04; a wallet
found there is a finding to stop on, not an auto-migration trigger.
PSBT is already solved by LND and already implemented: lnd.create-psbt
(WalletKit FundPsbt) and lnd.finalize-psbt (finalize + broadcast), both
rate-limited, on LND v0.18.4-beta. KEY-03 becomes: delete the Core path and
make that flow first-class, tested and documented, including that the PSBT
carries the BIP-32 key-origin data a hardware signer needs.
Records the standing honesty constraint that Lightning channel/revocation/HTLC
keys are not air-gappable at all, and defers the BDK+ElectrumX cold vault to
its own phase (D-07c).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Bitcoin Core's wallet is legacy and unused: bitcoin.init-wallet-from-seed has
no caller outside its dispatcher registration, the wallet UI is LND-only
(lnd.sendcoins/estimatefee/getinfo), archi-dev-box has no bitcoin/wallets/ dir
so the handler's named descriptor wallet was never created there, and the
endpoint is authenticated + password-gated so F-13 was never remotely
reachable.
F-13 is therefore latent, not live. D-07's migration premise is unproven, so
KEY-03 is re-scoped discovery-first: check the fleet for any wallet this
handler created before planning any migration. Migration + checkpoint stay,
conditional on discovery.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Reactions, replies, read-receipts, edits, deletes, forwards and channel sends
shared one bare `{ ok: true, sent: true }` case, so none of them rendered on
the demo — the UI derives reaction chips and reply quotes from the message
store, and there was nothing in it to derive from.
Each now mirrors its daemon counterpart. Reactions/replies/receipts push typed
messages carrying the { sender_pubkey, sender_seq } target key Mesh.vue's
reactionIndex and replyTargetPreview read. Edits rewrite the text and set
edited_at; deletes tombstone IN PLACE (plaintext, typed_payload.deleted,
message_type 'delete') because that is what mesh/mod.rs apply_local_delete
does — it does not remove the row.
Edits and deletes go through a per-session overrides overlay keyed by
sender_seq, because mesh.messages rebuilds its seed array on every read, so
in-place mutation would only ever work for messages sent this session.
mesh.refresh and mesh.reboot-radio stay acknowledgements on purpose — the
daemon's handlers have no message-store effect either — with a comment saying
so, so a later reader does not "fix" them into divergence.
Also completes the phase bookkeeping for 01-02/03/11/12/13/14/15 and lands the
orphaned 01-12/01-14 SUMMARYs.
Verified: parity harness 17/17 live assertions; full frontend suite 102 files
/ 822 tests green.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Ten methods the UI calls had no case at all in mock-backend.js, so the demo
answered them with "Method not found" and the frontend swallowed it in a
try/catch — peer renaming, scheduling, clear-all, the assistant panel and two
federation actions were all silently inert on the demo.
Each new handler mirrors its daemon counterpart and cites the Rust source it
mirrors, per the house convention above mesh.transport-advice:
mesh.contacts-list/-save typed_messages.rs (contacts merged over peers by
pubkey_hex; absent params leave stored fields)
mesh.clear-all status.rs ({ status: "cleared" })
mesh.schedule/list/cancel assistant.rs + scheduler.rs ScheduledMessage
mesh.assistant-status/-configure assistant.rs (key-presence semantics)
federation.cancel-request handlers.rs (outbound+sent only, notify defaults
true) — plus an outbound seed request, since
without one the demo's cancel path was
unexercisable
federation.notify-did-change handlers.rs ({ notified, failed, results })
mesh.peers and mesh.contacts-list now share one DEMO_MESH_PEERS list so they
cannot disagree about who is on the mesh, and the peer with no pubkey_hex is
omitted from contacts exactly as the daemon omits it.
scripts/mock-rpc-parity.mjs cross-references UI call sites against mock cases
and then drives a live scripted RPC sequence against an ephemeral-port
instance (MOCK_BACKEND_PORT). It matches only `method: '<x>'`, NOT bare string
literals: Mesh.vue and Federation.vue use the same dotted names as
resource-cache keys, and matching those would report permanent phantom gaps.
Fail-first proof: disabling the mesh.clear-all case makes the harness exit 1
naming the gap; restored, it exits 0 twice in a row with no stray listener.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Both paid-tick surfaces now show the EQ-segment ring instead of a CSS ripple
burst (send modal) and a plain circle (scan modal), via a new `badge` size
variant at 160px/192px with matching --viz-radius. A transform scale of the
compact variant was ruled out in 01-UI-SPEC.md because it would scale segment
stroke width and blur along with the geometry.
Both surfaces use the identical composition — a badge-sized relative container
with the checkmark core absolutely centred over the ring — so the two ticks
cannot drift apart visually.
ScreensaverRing also gains the prefers-reduced-motion guard it never had, on
the component rather than per call site, so the screensaver and
SystemDangerZone variants are covered too.
Also records live-browser verification for 01-13's scroll cue (three
viewports, plus a geometry probe of the hide condition).
Verified: 5 new tests; full suite 102 files / 822 tests green; npm run build
clean with viz-ring-badge present in the built bundle. The live visual
no-clipping observation Task 2 asks for is explicitly NOT done — recorded as
deferred to plan 01-07's consolidated sign-off.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
bip39::Mnemonic::generate(24) resolves through Mnemonic::generate_in to
&mut rand::thread_rng() INSIDE the bip39 crate (bip39-2.1.0/src/lib.rs:
311-313 -> :296-298 -> :267-283), so the entropy source behind Archipelago's
entire key hierarchy -- node Ed25519 did:key, node Nostr key, FIPS mesh key,
per-identity keys, the BIP-84 wallet, the LND aezeed entropy, and the fleet
release-root SIGNING key -- was chosen by a dependency default rather than
stated at the call site.
Not a vulnerability today: rand 0.8.5's thread_rng is a fork-protected
ChaCha12 CSPRNG seeded from getrandom(2). But it is precisely the structural
shape of the 2026-07-30 COLDCARD entropy defect (T1), where a refactor
silently rebound seed generation to a non-cryptographic PRNG with no compile
error and no test failure.
- New private helper generate_mnemonic_with<R: CryptoRng + RngCore> calls
bip39's injectable generate_in_with; MasterSeed::generate passes OsRng
explicitly, with the rationale pinned in a doc comment
- mnemonic_generation_uses_injected_rng: drives generation from a
deterministic test RNG and asserts the result equals from_entropy(exactly
the bytes that RNG emitted) -- direct proof the INJECTED rng is consumed --
plus a known-answer pin and a determinism check. This test cannot be
written against the previous code: there was no seam to inject through
- mnemonic_generation_is_256_bit: the OsRng path yields 24 words and two
successive productions differ
No change to derivation paths, word count, the empty-BIP-39-passphrase
decision, or the at-rest encryption envelope.
Verified: CARGO_INCREMENTAL=0 cargo test -p archipelago seed:: -> 25 passed,
0 failed.
Full analysis: docs/security/ENTROPY-SEED-AUDIT-2026-07-31.md (F-02, §4, §7).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
- Adds F-13 (High) to the audit: the BIP-84 account PRIVATE key is imported
into Bitcoin Core (bitcoin.rs:203 disable_private_keys=false, :229-231
wpkh(xprv/...)), so the spending key is persisted outside the Argon2
envelope in a wallet with an empty passphrase; the descriptors also carry
no [fingerprint/derivation] key origin, so no hardware signer could ever
use them. Found by tracing secret class (1) end-to-end
- Fills the Remediation Backlog: R-00..R-15, prioritised severity x effort,
each with the finding it closes, files, effort, and hardware gating; plus
an explicit "not implemented here, and why" section
- Records ARCHY-1 as APPLIED with the exact test evidence and an honest note
that making the source explicit removes a future failure mode rather than
repairing a past one
- Wires the resulting open items into docs/UNIFIED-TASK-TRACKER.md in its
existing tier/checkbox format: Tier 0 (cargo audit/deny CI, ceremony
mnemonic input, a five-item hygiene batch, secrets.rs OsRng), Tier 1 (ISO
fail-open first-boot secrets, Argon2 vs ADR-005, the on-node checklist),
Tier 2 (the Critical unauthenticated seed RPCs, PSBT Phase 1, PSBT phases
2-7, seed-RPC transport confinement)
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
On a short viewport the seed-confirmation tickbox sits below the fold inside
the step's scrolling area while Continue stays pinned and disabled in the
fixed footer — onboarding reads as broken rather than incomplete.
The cue is a sticky-bottom scrim and glass pill inside the scroll region, and
its visibility comes from real geometry: scrollHeight vs clientHeight for
overflow, then a getBoundingClientRect comparison of the tickbox's bottom
against the container's. On a tall screen the element does not render at all,
so those screens are unchanged. Rects rather than offsetTop because offsetTop
is relative to the nearest positioned ancestor — here the outer card, not the
scroll container.
It is wayfinding only: activating it scrolls the tickbox into view and never
sets confirmed, focuses Continue, or auto-ticks, which a test pins.
Listener setup was moved onto both onMounted paths — the sessionStorage
restore path returned early, so a user navigating back would have had no cue.
Verified: 6 new tests plus the full frontend suite green.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Entering picture-in-picture read as the lightbox being dismissed, and the
session died with it: both Teleport and KeepAlive move their subtree on
deactivation, which the PiP spec treats as removal.
MediaLightbox now listens for the video's own enterpictureinpicture event —
so PiP entered by the browser's native control behaves identically to the
toolbar button — and follows a fixed order: adopt, animate, then emit close.
Adopting first is what makes the element survive the unmount the emit
triggers; the invariant is documented in place so a refactor cannot reorder
it innocently.
The backdrop animates a handoff on the PiP path only, closing on transitionend
with a bounded 350ms fallback for browsers that skip the transition and for
the reduced-motion path where the duration is zero and the event never fires.
Verified: 5 new tests plus the full frontend suite green.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Six code paths configured the Lightning gateway with a bcrypt hash committed
to this repository — and one deploy path with a plaintext password literal —
whenever the per-install secret was missing. Anyone holding a copy of the repo
held the admin credential for every gateway that ever took a fallback.
container::secrets now owns the credential end to end: ensure_gateway_credential
(idempotent, delegates to ensure_one's bcrypt arm) and gateway_bcrypt_hash,
which returns Err when the secret is missing/empty and when the stored value is
on the KNOWN_DEFAULT_GATEWAY_HASHES denylist — so this codebase cannot hand
back the compromised value even to a node already carrying it.
get_app_config was widened to Result so a credential-less install cannot reach
podman run at all; configure_fedimint_lnd takes the resolved hash instead of
re-reading with its own fallback. The four shell paths stop generating
credentials entirely (dropping the htpasswd host dependency) and skip container
creation with a printed reason rather than substituting anything.
Naming converges on the manifest's fedimint-gateway-hash/.pw, with legacy
fedimint-gateway-password values copied forward rather than regenerated so no
node loses a working unique credential. Plan 01-16 owns rotation of installs
already carrying the default.
Verified: cargo build clean; cargo test -p archipelago 999 passed (2
boot_reconciler timing tests failed under concurrent load, green in isolation,
untouched by this diff); bash -n clean on all five scripts; the compromised
literal now appears exactly once in the tree, as the denylist.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Watch-only descriptor wallets, external signers, multisig, air-gap transport
and honest LND limits — a spec a future /gsd-plan-phase can consume.
- Names the current gap: bitcoin.rs:203 passes disable_private_keys=false and
bitcoin.rs:229-231 imports wpkh(xprv/...), so Core holds the BIP-84 account
PRIVATE key today. Closing that is Phase 1 and unblocks everything else
- Full Core RPC loop with wallet- vs node-scoped RPCs; analyzepsbt drives UI
- Tier 1 single-sig with mandatory [fingerprint/derivation] key origin; Tier 2
wsh(sortedmulti) on BIP-48; taproot/MuSig2 deferred as UNVERIFIED
- BC-UR v2 primary (fountain-coded, degrades gracefully), BBQr for Coldcard,
file fallback always; animated multi-frame is mandatory, not optional
- LND: channel/revocation/HTLC keys CANNOT be air-gapped; funding tx must
NEVER be self-broadcast (type-level refusal, not a boolean)
- On-chain (PSBT-protectable) vs lightning (necessarily hot) split, with the
exact user-facing sentence the UI must use
- Hot wallet kept as explicitly-secondary with server-enforced limits; safe
path is the DEFAULT, per T1's survivors
- Migration section refuses to over-alarm: the audit found no entropy defect,
so no Archipelago user needs to rotate a seed
- 7 phases with dependencies, candidate requirements, and hardware gating
- Answers two open items in docs/hardware-signer-design.md
- Flags bitcoin-knots:latest as an unpinned tag vs ADR-009
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Evidence-backed audit of every secret class against the real tree, prompted
by the 2026-07-30 Coinkite COLDCARD entropy incident.
- No Coldcard-class entropy defect exists: no non-cryptographic PRNG, no
clock-seeded key, no Math.random() in any browser key path
- F-01 (Critical, NOT entropy): seed.generate/seed.restore are unauthenticated,
unrated, and unconditionally overwrite a live node's Ed25519/Nostr/FIPS keys
- F-02 [ARCHY-1] CONFIRMED: mnemonic entropy source is a bip39 transitive
default, not a call-site argument — the exact structural shape of T1
- F-03 (High) [ARCHY-3]: first-boot TLS/SSH regeneration is fail-open and its
completion marker is set even on failure, over a fleet-shared cached rootfs
- ARCHY-2 confirmed good; ARCHY-5 refuted as a present defect (32 | 256)
- Argon2::default() is 19MiB/t=2, not ADR-005's stated 64MB/3
- Corrects the scoping assumption that image-recipe/_archived/ is dead: it is
the live ISO builder, exec'd by build-debian-iso.sh
- Adds a "What we do right" section and an UNVERIFIED on-node checklist
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
npm run build's vue-tsc -b pass caught TS2532 (possibly-undefined array
access) on two array-index reads the vitest run alone doesn't type-check —
optional-chain them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Following the confirmed 2026-07-30 Coinkite COLDCARD low-entropy incident,
plan three deliverables as a single 3-task quick plan:
1. docs/security/ENTROPY-SEED-AUDIT-2026-07-31.md — evidence-backed audit of
every key-material path (Rust/TS/shell), adjudicating research findings
[ARCHY-1]..[ARCHY-4] with file:line evidence, incl. the one-ISO-many-nodes
correlation risk and an explicit UNVERIFIED on-node checklist.
2. docs/security/PSBT-SIGNING-ARCHITECTURE.md — descriptor watch-only,
wsh(sortedmulti) multisig, air-gap transport, honest LND limits
(channel/revocation/HTLC keys cannot be air-gapped), hot wallet as
explicitly secondary, migration path, phased rollout.
3. Remediation backlog into docs/UNIFIED-TASK-TRACKER.md + one gated
hardening fix (explicit-OsRng injection at the mnemonic call site) proven
by a known-answer test that cannot exist before the change.
Audit-and-spec only — no wallet/signing implementation. Verify gates enforce
file:line evidence density and a secret-shaped-string check on both docs, and
assert no commit authored by this plan touches the concurrent agent's files.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Dorian verified the sibling-height match, internal scroll, and unchanged
stacked layout on his running dev session — all correct. The only issue was
the xl:min-h-[20rem] floor (an unmeasured judgement call, flagged as such in
01-12-PLAN.md): when node discovery is disabled, Web5NodeVisibility renders
short, the floor takes over, and 20rem left the Connected Nodes card looking
stunted. Doubled to xl:min-h-[40rem] per his direct instruction ("twice as
tall"). Test updated to pin the new value.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Task 2 of 01-01-PLAN.md, closing the gap left after Task 1's initial commit
(2f99db5e):
- record_peer_transport and update_node now hold FEDERATION_STORE_LOCK for
their whole load-mutate-save cycle via the *_inner variants, instead of
calling the public (separately-locked) load_nodes/save_nodes — closing
the same class of race the lock was introduced to fix, just for the two
mutators Task 1 didn't reach.
- Add test_remove_errors_when_tombstone_write_fails: pre-creates the
removed-nodes path as a directory so the tombstone write fails, then
asserts remove_node returns Err AND load_nodes still contains the node —
proving a failed removal never half-applies.
cargo test -p archipelago federation::storage: 14/14 green (was 11, +3 across
Task 1/2). cargo build -p archipelago: no new warnings, no dead-code warnings
on any *_inner fn. Public signatures unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Cloud.vue's viewPaidItem() called window.open() instead of the in-app
MediaLightbox, and its content.owned-get fetch had a 60s timeout with no
loading indicator and a swallowed catch. Moves the fetch/decode/route logic
into a new usePaidItemViewer composable: image/video route to a second
MediaLightbox instance fed a synthetic FileBrowserItem, audio still goes to
the global bottom-bar player, and anything with no in-app viewer keeps
today's browser-tab fallback. The Paid Files row now shows an "Opening…"
spinner (matching PeerFiles' existing treatment) for the fetch's duration,
becomes non-interactive to prevent double-fetch, and a real error surfaces
through the view's existing alert-error block instead of an empty catch.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
usePipSession() owns an off-screen div under document.body; adopt(video)
moves the element there so it survives the unmount of whatever view
rendered it (a Teleport and a KeepAlive'd view both move their subtree on
deactivation, which the picture-in-picture spec treats as removal).
release() tears down playback and detaches the element; a
leavepictureinpicture listener on the adopted element is the primary
release path so an orphaned owner can never leak it.
Also adds isPipSupported() to pip.ts — a call-time version of the existing
import-time pipSupported const, needed because a test can't restub
document.pictureInPictureEnabled after import. togglePip and pipSupported
are unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The three tab panes carried max-h-72 xl:max-h-none, so at the xl breakpoint
(where the Web5 row becomes two grid columns) the cap lifted with nothing to
replace it: the visible pane grew to fit every row, stretched the grid row,
and the scrollbar the user expects never appeared.
Give each pane xl:flex-1 xl:basis-0 xl:max-h-none instead — zero flex-basis
means the pane contributes no intrinsic height, so the grid row is sized by
the Web5NodeVisibility sibling alone, grid's default align-items: stretch
gives the card that height, and flex-1 hands the leftover height back to the
pane, which scrolls inside it via the existing overflow-y-auto. The card root
gets min-h-0 (so the flex column can shrink below content height) plus an
xl:min-h-[20rem] floor so a short sibling still leaves a usable list area
instead of collapsing to the header+tabs strip.
Below the row breakpoint nothing changes: the stacked cap (max-h-72) and
scroll are untouched, and dropping flex-auto (replaced by nothing, i.e. the
default 0 1 auto) has no visible effect since single-column stacked cards
have no extra flex space to distribute anyway.
Adds Web5ConnectedNodesScroll.test.ts to pin the contract across all three
tab panes and the card root so a future cleanup cannot reintroduce the
grow-to-fit regression a third time.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Release and OTA assets are live; the installer ISO was blocked three times by
a dirty shared tree. Records the exact command, TMPDIR requirement, the
background-execution lesson, the build-from-HEAD decision and its reasoning,
and which gate stages were already observed passing.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
apps/botfights/manifest.yml went to 1.2.11 in aea17248, but the two unsigned
catalogs (app-catalog/catalog.json and its neode-ui/public copy) still
advertised 1.2.9, failing the release gate's catalog-drift check and blocking
the ISO build. releases/app-catalog.json was already correct and signed.
Regenerated via scripts/generate-app-catalog.py (syncs from manifests, no key
needed). app_ports.rs was rewritten by the same generator; verified the port
set is byte-identical in content (35 ports, none added or removed) and
re-normalised with cargo fmt.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Ships botfight repo commits f5f57e6, c162d5e, 41f1b93, 10d4209:
- Fixed 6 instances of the same trust-a-client-supplied-pubkey pattern
across auth/payments/queue routes, two of them critical: GET
/winnings/:botId had no auth at all and leaked live spendable Cashu
bearer tokens; POST /connect-wallet let anyone redirect a victim
bot's future payouts to an attacker's wallet by pubkey (public by
design in nostr).
- Added owner-reachable AI-answer settings (existing bots, not just at
creation) and a "claim your winnings" UI (Cashu payouts were minted
server-side but had no frontend consumer at all until now).
Co-Authored-By: Claude <noreply@anthropic.com>
Both the 20G sparse qcow2 test disk and the serial console log were
hardcoded to /tmp regardless of $TMPDIR, so pointing the ISO release
build at a disk-backed scratch dir (to avoid tmpfs space pressure)
would not have covered this stage. Falls back to /tmp when TMPDIR is
unset — no behavior change for existing callers.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed by Dorian via scripts/sign-manifest.sh; signature verified
against the pinned release-root did:key by the signing script itself
and independently re-checked here (check-release-manifest.sh: version,
changelog line count, and both components' sha256/size all match the
on-disk artifacts).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
create-release-manifest.sh's changelog extraction pulls every non-blank
line between the version header and the next "## ", not just "- "
bullets — so the previous commit's "### Known gap" markdown heading and
its paragraph leaked into releases/manifest.json (and, via
sync-whats-new.py, the Settings "What's New" modal) as a malformed,
truncated entry (the closing clarification sentence was cut by the
extractor's 10-line cap).
Rewritten as a single "- " bullet, matching every other CHANGELOG entry,
so it renders cleanly and completely in both the OTA manifest and the
in-app modal instead of showing raw "### " syntax to node operators.
Also folds in core/Cargo.lock's version bump, which create-release.sh's
own commit step omits from its `git add` list.
Same binary/frontend artifacts as the prior commit (identical sha256/
size in the regenerated manifest) — only the changelog text changed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
scripts/sync-whats-new.py --check (part of the release gate) requires
every CHANGELOG version to have a matching block in the Settings
"What's New" modal. Also strips CHANGELOG markdown bold/italic markup
from the v1.7.119-alpha bullets first — the modal renderer only
strips backticks, not **/* emphasis, so it would have leaked literal
asterisks into the user-facing modal text.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Whitespace-only reflow in storage.rs/seed.rs/update.rs (rustfmt line-
wrapping rules) and app_ports.rs (array literal reflow after the port
list grew). No logic change. tests/release/run.sh's cargo-fmt --check
stage was failing on this before v1.7.119-alpha could be cut.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Written ahead of create-release.sh, which validates a curated
CHANGELOG.md section already exists rather than generating one.
Leads with FED-08 (private-channel invoice route hints) and FED-09
(doctor/Tor setgid restart-loop fix), covers phase-02 sessionStorage
security hardening and the PWA auto-update decision, and records the
5x lifecycle gate omission (node .228 unreachable) as a known gap
rather than a footnote.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Untracked Vite cache dir was tripping build-iso-release.sh's
git-status-porcelain clean-tree preflight check.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Pure npm metadata: 'peer': true flags dropped by a differing npm version.
No dependency added, removed, or version-changed. Committed rather than
reverted so nothing another session did is discarded; the ISO release
preflight requires a clean tree on main.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Phase 02 (ui-performance) verified passed. 12 plans (8 planned + 4 gap
closure). PERF-01/02/03 all met. Two residuals formally overridden by the
user: Discover's entrance-animation replay (animations are his domain) and
OpenWrtGateway (no device connected, unmeasurable).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Reassessed after the OpenWrtGateway override (commit b4350e24) closed the
single open item from the prior human_needed pass. Judged the override on
its merits: well-formed, corroborated by a direct first-person quote from
Dorian this session, with one noted imprecision (the rationale slightly
overstates that no measurement at all is obtainable, when the
disconnected-state UI could technically still be re-measured) that doesn't
change the substance of a legitimate stakeholder scope call. With both
residual, non-poller-fixable costs (Discover's animation replay,
OpenWrtGateway's untestable hardware dependency) now individually accepted
by rationale-backed override, and every other named regression fixed or
substantially recovered, status is passed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
No OpenWrt device is connected to this node, so the surface cannot be
exercised and no post-fix measurement is obtainable. The user decided to pass
it for this milestone. Also notes that its earlier baseline/regressed figures
measured a disconnected-device UI rather than the real screen, and records the
residual risk for the next milestone.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Independently re-derived all 02-11/02-12 claims: recomputed every claimed
median directly from 02-PERF-FINAL.json's raw samples (all match), confirmed
the frozen perf harness is still byte-for-byte untouched, re-ran the
keepAliveLifecycle regression suite (19/19) and the full workspace suite
(788/788), and confirmed the three leaked-poller fixes are genuinely wired
in source. Discover's override is well-formed and corroborated across three
independent artifacts. Status: human_needed — OpenWrtGateway is the one
named regression with zero post-fix measurement (crashed for an unrelated
reason) and no override scoped to it; everything else is either verified
fixed, substantially recovered with an honest residual, or formally
overridden by Dorian.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Dorian decided entrance-animation behavior is his domain and is not to be
changed, cancelling this gap-closure plan before any source file shipped.
The composable drafted during investigation was deleted (never imported).
Preserves the investigation findings (blast radius, the ~241ms-1716ms
measured cascade, the first-paint-901ms-of-1095ms evidence, the Apps-revisit
184-267ms calibration, and the pre-phase-2-baseline note) so the analysis
isn't lost or re-done. Discover's revisit cost stands as a formally accepted
deviation at its 02-11 FINAL number.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signs releases/app-catalog.json after the corrected manifest fix in
3c7a1fbb (data_uid: 999:999, matching the image's actual internal UID).
Co-Authored-By: Claude <noreply@anthropic.com>
The re-verification required either a fix or an accountable stakeholder
decision. Plan 02-11 fixed four of five measurable surfaces. 02-12 isolated
Discover's remainder to the card-stagger entrance animation replaying on every
revisit; the only fix changes animation behavior, which Dorian rules off-limits
as his domain. Decision recorded with rationale, evidence and attribution.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The prior fix (5745db51) copied the fedimint-clientd/barkd data_uid
pattern (1001) without verifying it against this image. Live on
x250-beta the container still crash-looped with the same SqliteError:
unable to open database file — `podman exec botfights id` showed
uid=999(botfights) gid=999(botfights), not 1001. The image's
Dockerfile does `useradd --system` with no explicit UID, which lands
at 999, and security.user in the manifest is descriptive only — it is
not read by this app's (non-Quadlet) install path, so it can't be used
as the source of truth either.
Fixed data_uid to 999:999 and corrected security.user to 999 to match
reality. Verified live: manually re-chowned the existing bind mount to
999:999 on x250-beta, restarted the container, confirmed
`database migrated` + `listening on http://localhost:9100` +
`/api/health` returns ok, and confirmed /api/bots on x250-beta returns
identical data to the canonical arena (arena-proxy forwarding
correctly). Regenerated catalog; cargo test -p archipelago-container
manifest passes (38/38).
Co-Authored-By: Claude <noreply@anthropic.com>
Signs releases/app-catalog.json after the manifest fix in 5745db51
(absolute bind-mount path + data_uid: "1001:1001" for botfights).
Co-Authored-By: Claude <noreply@anthropic.com>
Two root-caused bugs found live during a fresh install on a second node
(x250-beta) that never surfaced on archi-dev-box by accident of that
node's prior state:
- volumes.source was a bare relative "botfights-data" instead of an
absolute host path, inconsistent with every other app's manifest.
Resolved to /var/lib/archipelago/botfights on archi-dev-box but to
/home/archipelago/botfights-data on x250-beta, which doesn't exist
there. Fixed to the absolute path, matching netbird-server and every
other app's convention.
- data_uid was missing entirely. The container runs as internal UID
1001 (security.user), but without data_uid the orchestrator's bind-dir
ownership fixup only fires via a same-owner-as-anchor fallback that
assumes no-data_uid apps run as container-internal root. Root cause of
a real SqliteError: unable to open database file crash-loop on
startup. Fixed by adding data_uid: "1001:1001", same pattern as
fedimint-clientd and barkd.
Bumped catalog via generate-app-catalog.sh; cargo test -p
archipelago-container manifest passes (38/38).
Co-Authored-By: Claude <noreply@anthropic.com>
releases/app-catalog.json regenerated and signed by the release-root key
(did:key:z6MkkidEnEpo6qHMCNSZoNKWtvQvxq3whnaME9wGgEFhq7ur). botfights is
the only content change vs the previously-published catalog (verified
structurally — all other 65 apps unchanged): 1.2.8 -> 1.2.9.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Bumps to botfights:1.2.9, which carries: Cashu token payment wired as the
primary entry-fee UX in WalletConnect.vue (was built server-side already
but never called from any UI), Lightning/NWC demoted to secondary, and a
fix so anonymous poll-mode bots (not just nostr-authenticated humans) can
use ranked/staked fights — join-ranked previously required a pubkey
unconditionally, silently locking out the entire AI-agent audience.
No archy-side manifest changes beyond the version/image bump.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
02-11 named a second, distinct cause of Discover's revisit slowness with
full profiling evidence (card-stagger/home-card-animate classes never
removed from the DOM, so Chromium restarts the CSS entrance animation on
every KeepAlive reattach) but deliberately left it unfixed — the blast
radius (Discover/Apps/Marketplace/Home + Web5 sub-cards) exceeded that
plan's scope. This gap plan fixes every affected site with one shared
composable rather than per-file patches, decouples Home.vue's dual-purpose
animateCards ref so its overlay/EasyHome visibility logic stays untouched,
and ends at a blocking human-verify checkpoint before any SUMMARY is written.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
fix_tor_permissions() exact-matched stat's output against the literal
string "700", but stat -c '%a' omits leading zeros so Tor's own
setgid HiddenServiceDir mode (2700) never matched. Every ~5-minute
doctor run "fixed" it back to 700, restarted Tor, and Tor immediately
set 2700 again — so Tor never survived long enough to build a usable
consensus/HSDir cache, breaking the mesh's Tor fallback entirely.
- Compare only the last 3 mode digits (owner/group/other), which is
the property that actually matters, so 700 and 2700 both pass while
750/707/2755 etc. are still corrected.
- Add a 30-minute restart backoff (timestamp file under
/var/lib/archipelago/) so no future condition can reproduce a
restart storm even if the fix fires repeatedly.
- Log clearly in both directions: a debug-level no-op line when a
directory is already correct, and an explicit "was NOT fully
denied" line when a real fix is applied, plus a line when a restart
is skipped by the backoff.
Verified statically against temp directories (2700 accepted with 0
restarts; 750/707/2755 corrected with exactly 1 restart; a second
real fix inside the backoff window logs a skip instead of
restarting). No live Tor/doctor/systemd unit was touched.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The code-fixer correctly declined this change twice on relayed consent. It was
then made by the orchestrator holding first-hand authorization from Dorian, who
chose forced auto-update with the mid-payment reload risk explicitly stated and
rejected warning UI. Records the quote, and that the implementation reused the
existing kiosk auto-apply path (preserving the cinematic and first-install
guards) rather than flipping build-time skipWaiting.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Phase 02's last open item closed: web5/server/fleet fixed and proven
on archi-dev-box; app-details restored to at/near baseline;
discover's second cause (CSS entrance-animation replay on KeepAlive
reactivation) named and evidenced but not fixed, scoped as a
follow-up; openwrt-gateway not measurable this pass, prior numbers
flagged not retracted.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
container-doctor.sh's fix_tor_permissions() exact-matches mode '700', but Tor
sets its hidden-service dirs to 2700 (setgid). Every 5-minute doctor run
'fixes' 2700->700 and restarts tor@default; Tor resets it and the cycle
repeats, so Tor never builds a usable consensus/HSDir cache. Result on a live
node: 'No more HSDir available to query', onion peers unresolvable, and with
the FIPS direct path also timing out, mesh sends failed entirely.
Planned as 01-20 in wave 1 so it ships in the same release. FIPS direct
connect_fail is a separate concern, handed to FED-03's transport review.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Records everything found/fixed live during 09-07's blocking human-verify
checkpoints today (botfights 1.2.2-1.2.8): iframe embed, native signer
bridge, mode-picker discoverability, proxy-URL leaks, round-jump backfill,
broken profile images (CSP), AI-answer discoverability, and the webhook_test
signature exception. Also records the full threat register worked through
for the user-directed Cashu fixed-stake entry-fee request (21 sats, winner
takes all) — mint configured and verified live (Minibits), actual
token-accepting implementation deliberately NOT done yet, scoped as a
follow-up.
Note: this section was written once already earlier in the session and
appears to have been lost to a shared-tree overwrite before it was
committed (CLAUDE.md "Concurrent agent in shared tree" hazard) — recreated
and committed immediately this time.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Deployed the poller fix to archi-dev-box and re-ran the frozen harness
(02-PERF-FINAL.json, 5 runs/surface). Web5 fixed (275ms, below both
its 566ms pre-phase-2 baseline and the 300ms pass bar). Server's
regression closed (574ms, below 738ms baseline) though not yet under
300ms. Fleet substantially improved (790ms, down from a 2631ms
regression). AppDetails restored to at/near its own baseline.
Investigated the two open items the coordinator raised:
- OpenWrtGateway: this run's 5/5 samples failed with a Chromium
"Target crashed" error cascading from an unrelated surface earlier
in the same harness run — recorded as not-measurable, not written
in as data. Separately confirmed via source (OpenWrtGateway.vue's
h1 renders unconditionally, and a "No router configured" RPC error
deterministically shows a real Connect-to-Router form) that the
prior baseline/after/remeasure numbers were measuring a genuine,
substantive disconnected-state UI render, not an empty/error page —
so the "six confirmed regressions" count is not retracted, but the
numbers are flagged as reflecting one specific code branch.
- Discover (1389ms, worst remaining, least improved): profiled
directly and found a SECOND, distinct, phase-2-class cause —
showStagger/card-stagger entrance-animation classes are baked into
the DOM at first mount and never programmatically removed (the flag
is a correctly-scoped once-per-session const, but nothing ever
re-renders to strip the class), so every KeepAlive detach/reattach
cycle restarts the CSS animation on reactivation, replaying the full
entrance cascade on every revisit. Confirmed via a diagnostic
showing DOM card count doubling transiently on every revisit and an
extended animationstart/animationend event log. Not fixed this pass
— the safe fix's blast radius spans 5+ files outside this plan's
scope (Apps.vue, Marketplace.vue, Home.vue, several Web5 sub-cards)
and needs its own real-device verification budget, matching the
precedent 02-02's original KeepAlive rollout needed for this exact
class of change. Named and evidenced, recommended as a dedicated
follow-up rather than expanded into this plan under time pressure.
REQUIREMENTS.md's PERF-02/PERF-03 rows updated to the final state.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
releases/app-catalog.json regenerated and signed by the release-root key
(did:key:z6MkkidEnEpo6qHMCNSZoNKWtvQvxq3whnaME9wGgEFhq7ur). botfights is
the only content change vs the previously-published catalog (verified
structurally — all other 65 apps unchanged): 1.2.7 -> 1.2.8, fixes broken
profile pictures (CSP img-src) and makes the AI-answer feature discoverable
by default.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Bumps to botfights:1.2.8, which carries: the CSP img-src fix (nostr profile
pictures come from user-supplied kind:0 metadata URLs on arbitrary domains
— img-src was locked to 'self'/data:/blob: with no https:, so every
external profile picture rendered as a broken image), and discoverability
fixes for the new AI-answer feature (poll mode is now the default
connection mode, matching BOTFIGHTS.md's own documented default, and the
AI-answer section is expanded by default instead of collapsed behind an
extra click).
Canonical arena on VPS2 already rolled to 1.2.8 directly (docker compose
pull/up) ahead of this catalog publish, since it's a separate deployment
from the archy-catalog-driven per-node install path.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fleet.vue's useFleetData() (60s telemetry.fleet-status/-alerts poll),
Server.vue's FipsNetworkCard.vue (15s fips.status poll), and Web5.vue's
Web5Monitoring.vue (30s system.stats poll — redundant with Home.vue's
own correctly-gated 10s poll of the same store) all armed their
setInterval in onMounted and only disarmed it in onUnmounted/
onBeforeUnmount. That was harmless before 02-04 registered their
owning views in KEEP_ALIVE_PATHS (the view was destroyed on every
tab-away, so the teardown hook fired every time); once KeepAlive keeps
the instance alive, the teardown hook never fires again and the poll
ran forever in the background regardless of which dashboard tab was
showing.
Gated arm/disarm to onActivated/onDeactivated, mirroring Server.vue's
own vpnPollInterval fix from 02-04 exactly. Added regression tests to
keepAliveLifecycle.test.ts mounting each real component under a
synthetic KeepAlive with fake timers; confirmed RED against the
pre-fix code (git stash) before confirming GREEN with the fix restored.
Full suite (95 files/788 tests), type-check and build all green.
keepAliveTabs.test.ts is byte-for-byte unmodified.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
handle_lnd_createinvoice posted to LND's /v1/invoices with only value/memo,
so LND defaulted private to false and returned invoices with empty
route_hints. Any node whose only channels are private/unannounced was
unpayable through the wallet UI's Receive flow. Diagnosed on
archy-x250-mad2, whose only channel (to Olympus by ZEUS) is private with
~40.8k sats usable inbound; three wallet-UI invoices never received an
HTLC.
Audited every other invoice-creation call site and found a second one with
the identical omission: create_invoice, the seller-side/peer-file paid-
content flow (content.rs -> handler for paid downloads). Same bug, same
fix, wider blast radius than the one-node report suggested -- paid-file
sales were unreceivable on private-channel nodes too.
Extracted both call sites' invoice_body construction into one shared
build_invoice_request_body() that sets private: true unconditionally, and
added a unit test pinning that field so neither site can silently drift
back to false. private:true is harmless on nodes with public channels --
LND still prefers a direct public route and the hint is just an unused
alternate path.
CPU profile evidence (CDP Profiler + Tracing, additive
neode-ui/e2e/perf/profile-revisit.spec.ts, frozen harness untouched)
shows 86-99% of every revisit window spent in (idle)/(program) with
under 10% genuine app JS self-time on every surface — ruling out
expensive computed re-evaluation, watcher cascades, and whole-subtree
re-renders as the dominant cost, per the plan's own explicit list of
hypotheses to check before accepting one.
A follow-up source-level lifecycle audit (grep every setInterval call
site for a missing onActivated/onDeactivated pair, extending 02-04's
own audit convention past the top-level view files it originally
checked) found three child components/composables inside the
KeepAlive'd Fleet/Server/Web5 subtrees that arm a poll in onMounted
and only ever clear it in onUnmounted — harmless before phase 2
(the view was destroyed on tab-away) and now a permanent, session-long
background-RPC cost once KeepAlive keeps the parent instance alive:
useFleetData.ts (60s), FipsNetworkCard.vue (15s, Server), and
Web5Monitoring.vue (30s, Web5 — redundant with Home.vue's own,
correctly-gated 10s poll of the same store).
This directly explains the idle-dominated CPU signature (background
network/scheduling contention, not compute) and why 02-10's 5-run
remeasure regressed further than the 3-run baseline/after runs even
as disk pressure eased: a longer session accumulates more of these
always-on pollers, degrading every subsequent navigation.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
releases/app-catalog.json regenerated and signed by the release-root key
(did:key:z6MkkidEnEpo6qHMCNSZoNKWtvQvxq3whnaME9wGgEFhq7ur). botfights is
the only content change vs the previously-published catalog (verified
structurally — all other 65 apps unchanged): 1.2.3 -> 1.2.7, carries the
nostr-provider.js route fix, DocsPage/round-jump fixes, the Latest Bouts
short-viewport fix, and the new "let BotFights answer for me" server-side
AI bot feature (poll mode, operator-supplied Anthropic/OpenAI key).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Bumps to botfights:1.2.7, which carries the botfight repo's new
"let BotFights answer for me" feature (ca5b634) — an operator can paste an
Anthropic or OpenAI API key so this node's BotFights server answers fight
challenges automatically for a poll-mode bot, no external script needed.
Storage/security follows the same pattern this node already uses for its
own AIUI Anthropic key (system.settings.set "claude_api_key" in
core/archipelago/src/api/rpc/system/handlers.rs): 0600 file, never echoed
back. Also carries the nostr-provider.js 404 fix, DocsPage/round-jump
fixes, and the HomePage "Latest Bouts" short-viewport visibility fix from
the prior 1.2.4-1.2.6 iterations that were built and tested locally but
not yet pushed through the signed-catalog path.
No archy-side manifest changes beyond the version/image bump — 1.2.2's
ARCHY_EMBEDDED/ARENA_UPSTREAM_URL/generated JWT_SECRET are unchanged;
the new AI-bot feature's key storage lives entirely inside the botfights
app's own data volume, no new archy secret/env wiring needed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fleet/AppDetails/Web5/OpenWrtGateway/Server/Discover all measured
slower on revisit than the pre-phase-2 baseline (02-10's three-way
dispersion analysis). This plan profiles the real cause per surface
before touching source (D-10), fixes what's fixable without
reverting any T-02-01 persist:false decision or touching the
visual/animation contract, and re-measures with the frozen harness.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
archy-x250-mad2 is a user's device holding real funds. Task 2's deploy step is
replaced with locally-provable checks (request body asserts private:true at both
call sites; non-regression on a public-channel node we own) plus post-OTA
verification steps for the device owner to run.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Diagnosed on archy-x250-mad2: handle_lnd_createinvoice omits LND's private
flag, so invoices carry route_hints: [] and are unroutable for any node whose
channels are unannounced. Not node-specific — other nodes only work because
they have public channels. Planned as 01-19 in wave 1 so it ships early.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Phase 1 success criteria 7-13 were added 2026-07-30, after the phase's
original 10 plans were written. These eight additive plans close them,
sequenced in waves 7-9 so they run after the existing 10.
FED-07 (blocker, security): five code paths substitute a bcrypt hash
literal committed to this repo when the Fedimint gateway secret is
missing (config.rs, dependencies.rs, first-boot-containers.sh,
deploy-to-target.sh, deploy-tailscale.sh), and one deploy path
substitutes a plaintext password literal. 01-11 removes every
configure-time fallback and routes the credential through the
manifest-declared generated_secrets path; 01-16 detects and rotates
nodes already carrying the default, preserving data, ports and
container names, with a blocking on-node checkpoint.
UIFIX-01..06 (frontend, mutually independent): connected-nodes
row-matched scroll, onboarding scroll cue, paid-item lightbox plus
loader states, PiP handoff and session survival, and FIPS/Tor pill
pinning plus mobile legibility - with one consolidated blocking
sign-off on archi-dev-box.
ROADMAP: phase 1 plan count 10 -> 18, new plans appended with waves.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
deploy-to-target.sh rsyncs with --delete to TARGET_DIR=/home/archipelago/archy,
which is a symlink to /home/archipelago/Projects/archy. archi-dev-box is this
same machine over loopback SSH, so deploying from the main checkout is a no-op
(source and destination resolve identically) — but deploying from a git worktree
nested under it mirrored that worktree onto the main checkout and deleted
everything else: ~1810 tracked files, the deploying worktree itself mid-run, a
running dev server, and two concurrent sessions' uncommitted work.
Guard compares /etc/machine-id across the SSH hop and, when source and
destination are on the same host, refuses if either path contains the other.
Identical paths still pass, so normal local deploys are unaffected.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Alpha-stage, user-approved: a prompt only reaches users who click it, so
security fixes sat unapplied in long-lived sessions (installed PWA, kiosk
displays). Extends the existing kiosk-only auto-apply to all non-demo
clients.
Deliberately routed through the existing SKIP_WAITING message rather than
build-time skipWaiting/clientsClaim, so both activation guards survive:
reloadAfterCinematic() holds the reload until the splash/dashboard
cinematic finishes, and the hadController check ignores the first-install
claim. A build-time skipWaiting would bypass both.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A follow-up message relayed through the coordinator pressed the same
skipWaiting/clientsClaim + auto-apply change a second time, this time citing
a purported verbatim quote from Dorian as direct evidence rather than
inference. Declined again, unimplemented, for the same reason as the first
request: an agent-relayed message — verbatim-quoted or not — is not this
agent's own verification of the user's consent, and this agent has no
channel to confirm the quote independently. vite.config.ts and
PWAUpdatePrompt.vue remain unmodified.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
- New `## Re-measurement (gap closure)` section in 02-FINDINGS.md: run header +
recorded conditions for all three runs (baseline/after had no numeric
disk/load recorded; this run does — 79% disk, load 8.70-11.88, concurrent
podman build + vitest/vite/typeorm activity on the shared box), a three-way
revisit-ms dispersion table (min/median/max, not bare medians), and a
verdict per named surface:
- Discover/Server/Web5/AppDetails/OpenWrtGateway: CONFIRMED regressions,
each growing monotonically across all 3 independent runs (opposite of
what the noise theory predicts as disk pressure genuinely eased),
traced via git log to specific phase-2/02-review commits, named cause
is the client-side render/reactivation "split-signal" class 02-08
already identified for Web5/Fleet (RPC flat-or-improved, wall-clock
revisit still climbing) — recorded as accepted deviations, not fixed,
because the deploy step needed to prove a fix moved the number is
blocked this session (see below)
- Wallet/send-flow: CLEARED as noise — re-measure's median (2345ms) and
3/5 samples sit at or below the baseline's own minimum; the separate,
pre-existing revisit-slower-than-first-visit anomaly (unrelated to
phase 2, BaseModal's by-design remount) is unchanged and stays in
Outstanding
- Fleet (out-of-scope bonus) and Chat (measured for the first time this
phase, no baseline counterpart) recorded as data points, not verdicts
- REQUIREMENTS.md: PERF-02/PERF-03 traceability rows updated to point at
this section (scope deviation from files_modified, per plan checker note
— the coverage-table update this plan's own Task 2 text calls for)
Deploy note: Task 1 deployed archi-dev-box to 3e3159fa (frontend-only,
clean tree, dirty=false) before measuring, since 4 of 6 named surfaces are
touched by the 02-review commits the previously-deployed 8fe6217b predates.
Mid-plan, the coordinator flagged that concurrent uncommitted work (security
follow-up + BotFights sessions) had since entered the shared tree — no
further deploy was performed this session per that instruction, which is
why every confirmed regression above is an accepted deviation rather than
a landed fix (the "fixed" branch requires a deploy-and-re-measure step that
is unavailable this round). `neode-ui/e2e/perf/02-REVIEW.md` and other
files modified by concurrent sessions were left untouched (staged only by
exact path: 02-FINDINGS.md, REQUIREMENTS.md).
Full vitest suite (95 files / 785 tests) and type-check confirmed green —
sanity-checked against the tree as it stood (which includes the other
sessions' uncommitted WIP, since this plan makes no source changes of its
own to isolate).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Appends a dated addendum to 02-REVIEW.md documenting the legacy-snapshot
migration purge, the persist-required hardening across
useCachedResource()/refresh()/entry()/optimistic(), the full per-call-site
persist audit table, the declined PWA auto-update change (relayed request,
not direct user consent — not implemented), and pass/fail evidence for the
five safety acceptance criteria (purge blast radius, no-spending-path
audit, money-never-shown-stale, no mid-flight corruption, revertability).
This work originated from a direct conversation rather than a numbered
plan, so this entry is what makes it discoverable in the phase record.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Now that persist is required (no default) on useCachedResource()/refresh(),
every call site that previously relied on the implicit persist:true default
needs an explicit decision. Full audit, decision rule: money/identity/
peer-identity payloads -> false; static/aggregate/non-identifying data ->
true; ambiguous cases fail safe to false and are called out below.
persist:false (financial / identity / peer-identity payload):
- LightningChannelsPanel.vue: lnd.channels, lnd.closed-channels (open/closed
Lightning channel balances — wallet data, same class as CR-01's lnd-info)
- Cloud.vue: cloud.paid-items (carries paid_sats + purchase history),
cloud.peer-nodes (PeerNode carries did/pubkey/onion)
- Cloud.vue/PeerFiles.vue: cloud.my-files — not a clean money/identity/
peer-identity case, but a private per-user file listing; chosen false as
the fail-safe default per the audit rule, flagged here for review
- Credentials.vue: credentials.identities, credentials.list
- Federation.vue: federation.nodes (FederatedNode carries did — matches
Mesh.vue's already-persist:false federation.nodes decision)
- FipsSeedAnchorsCard.vue: server.fips-seed-anchors (SeedAnchor carries npub)
- Server.vue + FipsNetworkCard.vue: server.fips-summary corrected from
persist:true to persist:false — this shared cache key's real fips.status
response carries npub (this node's own FIPS identity key), which
Server.vue's narrower local type didn't surface but FipsNetworkCard.vue's
fuller FipsStatus type does; both call sites must agree since a mismatch
trips the dev-only entry() persist-consistency warning. Found during this
audit, not part of the originally-scoped call-site list — corrected as a
same-class T-02-01 violation. serverTabCache.test.ts updated to match.
persist:true (aggregate/status/public data, no identity or money):
- AppDetails.vue: app-details:bitcoin-sync (block height/sync progress)
- Cloud.vue: cloud.section-counts (bare per-section item counts);
cloud.peer-browse (browsePeer()/loadCatalog()'s direct resources.refresh()
calls now pass { persist: true } explicitly, matching the pre-existing
decision already documented at peerBrowseEntry())
- Federation.vue: federation.dwn-status (sync status/counters only)
- MarketplaceAppDetails.vue: app-details:versions (public catalog metadata)
- Monitoring.vue: monitoring.current/history/alerts/alert-rules (system
metrics and alert metadata only)
- OpenWrtGateway.vue: server.openwrt-status (network/router status, matches
sibling server.* resources)
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
CR-01 fixed web5.lnd-info/web5.networking-profits to persist:false, stopping
FUTURE writes to sessionStorage, but a tab already open before the update
ships reloads in-place onto the new bundle and keeps whatever the OLD
bundle already wrote under the old decision — indefinitely, since nothing
but clearAll() (logout) ever purges a resource: snapshot. Long-lived tabs
(installed PWA, kiosk display) are normal here, so this left updating users
exposed to exactly the T-02-01 exposure CR-01 was meant to close.
- Add a schema-version marker (resource:__schema) checked once at store
setup: absent or stale marker purges every resource:-prefixed
sessionStorage key, then writes the current version. One-time per tab
session (a matching marker no-ops), not per navigation/reload, so this
doesn't defeat the instant-paint-from-snapshot benefit the cache exists
for. CURRENT_SCHEMA_VERSION must be bumped whenever a key's persist
decision changes, documented inline as the contract for future changes.
- Extract clearAll()'s purge loop into purgeAllSnapshots(), reused by both
clearAll() (logout, T-02-02) and the new migration, so there's one place
that enumerates/removes resource: keys.
- Close the residual refresh()/useCachedResource() default: opts.persist
?? true was the exact footgun that caused CR-01 (a call site silently
opting into persistence by omission). persist is now a required
parameter on refresh() and useCachedResource()'s options, matching the
entry()/optimistic() hardening WR-04 already applied.
- Tests: legacy snapshot (no/stale marker) is purged on init; a snapshot
under the current marker survives a later init (proves one-time, not
every-boot); persist:false never writes a snapshot; marker is written
after purge; purge is strictly bounded to the resource: prefix (seeded
non-resource: sessionStorage keys and a localStorage auth flag survive
byte-for-byte); migration cannot race an in-flight fetch (runs
synchronously at store setup, before entries/inflight can hold anything).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
