Verified the security subsystem's design-doc claims against code:
- KEY-05's foundational claims are accurate: entropy::draw_key_bytes exists,
KeyGenRng is sealed with OsRng as its sole production member, MIN_GUARDED_LEN
is 12, and core/clippy.toml bans rand::random/thread_rng exactly as stated.
- But its per-site table listed every production nonce/key site as disposition
"migrate" (pending), when all of them have since been migrated to
draw_key_bytes(OsRng) — storage_crypto, credentials/store, wallet/bdhke,
mesh/x3dh — and zero rand::random/thread_rng remain in production. Added a
completion note so the doc no longer reads as pending work.
Both KEY-05 and PSBT-SIGNING-ARCHITECTURE referenced
ENTROPY-SEED-AUDIT-2026-07-31.md five times as their evidence base — a doc that
was moved to local-only, so a public reader could not follow it. Reworded all
five to state the audit's findings inline ("the internal entropy audit found
...") without the unresolvable path. No published doc references it now. The
link-checker missed these because they were inline code, not markdown links.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
apply_data_uid ran a recursive sudo chown on every prepare_for_start, and the
reconciler re-prepares — archi-dev-box's journal showed postgres-btcpay rechowned
every ~45s despite already-correct ownership, and on framework-pt the same loop
surfaced as operator-visible 'chown failed' noise. chown_for_rootless_container
now stats the target first and returns early when the top-level owner already
matches the host-mapped uid:gid. Deep drift in a running container is still
caught by ensure_running_container_ownership's in-container write-probe, which
is the authority that actually matters (it probes writability, not stat bits).
Co-Authored-By: Claude <noreply@anthropic.com>
The guide walked a developer from manifest to install but never told them how
to validate the manifest locally first — despite scripts/validate-app-manifest.sh
existing for exactly that. A developer's first signal that their manifest was
wrong would have been an install failure on a node.
Adds a "Validate Your Manifest" step at the top of Testing, pointing at the
script (recently fixed — it had been rejecting every manifest because it shelled
out to a missing ruby). Notes the strict behaviour a new submitter hits, e.g.
an unpinned :latest tag is rejected, and that the Rust parser is canonical.
Verified: the install RPC example in this guide (id + dockerImage) matches the
handler; the cargo test target crate name (archipelago-container) is correct.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Verified all 144 documented RPC methods against the dispatcher. 143 are live;
one was fabricated: `mesh.discover` (params { timeout_secs? }, returns
{ nodes: MeshNode[] }) does not exist — "mesh discovery" appears only in code
comments as a concept, never as a method. A developer calling it gets "unknown
method".
Replaced with the real peer-listing method `mesh.peers` (no params, returns
{ peers, count }), which the frontend actually uses and which was undocumented.
Also verified: every source path cited across the docs resolves (placeholders
and a correctly-recorded deletion aside), and every documented app-manifest
field exists in the schema (no fabricated fields).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Three `MEMORY → <note>` see-also references pointed at the private agent-memory
system from public docs (demo-deployment-design.md x2, registry-manifest-design.md
x1). Removed. No tracked doc references the memory system now.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
bitcoin-multi-version-design.md carried three layers of stale internal content:
an 80-line HTML-comment work-tracking block (per-phase status with "UNCOMMITTED
on the branch", node numbers, "Next action when resuming", "Decisions still
needed from user"); a rendered "Status: design (2026-06-22)" header that was
wrong — the feature shipped, all four phases, with the downgrade guard added
today; two private `MEMORY →` references; and a node-numbered scheduling note.
Now: the comment block is gone, the status reflects reality, the MEMORY
references and node numbers are removed, and "verify on a real node" replaced
the specific fleet addresses. The design content (source-of-truth decision,
phase designs, invariants) is unchanged.
Separately, bitcoin-version-bulletproof-rollout.md was an inter-agent rollout
handoff — node numbers, branch coordination, "the other agent owns" — not a
design or reference doc. Moved to local-only (still on disk, gitignored) like
the other handoffs; its two path references (a plan doc and a script comment)
are generalized.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This 2026-04 plan has been implemented, but it still read as an active plan
("implementation started"), linked private agent-memory paths, and ended with a
stale "To resume" work block naming fleet nodes, dated fleet state, and the next
file to edit.
- Header now marks it a historical design record and points at
container-lifecycle.md for the current behaviour.
- Removed the two private ~/.claude/.../memory/ references from the header and
the entire "To resume" section (private paths, node numbers, 2026-04-22 fleet
snapshot — none of it belongs in a public design doc).
- Genericized the one remaining node-number reference in the incident narrative.
The valuable content — the six failure modes and the reconciler reasoning that
answered them — is kept intact.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The open-source plan flagged three references as "the real gaps for app
developers", and the docs index named them as not-yet-written. Written now,
each from the code rather than stubbed:
- secrets.md — generated_secrets/secret_env: the two halves, the four kinds
(hex16/hex32/base64/bcrypt) and which files each writes, the idempotent
self-healing 0600 materialisation, and the rules a developer must not break
(no hardcoded fallbacks, one canonical name, right encoding). From
container/secrets.rs and the manifest schema.
- quadlet-compilation.md — manifest -> .container unit: the full directive
mapping (including Secret= by reference, never value, and Pull=never), where
units land (~/.config/containers/systemd, systemctl --user), the
render/write/enable/disable lifecycle with write-if-changed, and how to
inspect one. From container/quadlet.rs, scoped accurately to the companion-UI
path it drives today.
- container-lifecycle.md — the level-triggered 30s reconciler: desired state
from user-stopped/user-uninstalled/manifest set, the operations table, the
self-heal-vs-respect-a-deliberate-stop rule, and migrations-never-destroy-data.
From prod_orchestrator.rs and boot_reconciler.rs.
Index updated to link all three under App development and the "known gap" note
removed. Every link across the docs tree resolves.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Follow-up to 73970cf3. Two improvements:
- Corrects a stale claim — manifests are no longer "loaded from disk, goal is
the catalog". The signed catalog has been the delivery mechanism since
2026-06-23 (origin-wins over disk), so the guide states that, plus the
consequence contributors need: editing a disk manifest alone does not change
a catalog-covered app.
- Tightens the north-star paragraph.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
CLAUDE.md was the internal agent guide: a dated "gate is GREEN" status banner
naming a specific node, pointers to now-local-only planning docs
(PRODUCTION-MASTER-PLAN, UNIFIED-TASK-TRACKER, multinode-testing-plan), the
gitea-ai push account mechanics, and references to the private memory system.
Rewritten as a contributor guide that keeps everything public-worthy — the
invariants (rootless podman, declarative apps, manifest-declared secrets,
non-destructive migrations), the build/verify notes, the commit-and-push
discipline, and the production test-gate definition — and drops the status,
node numbers, push-account specifics, and memory references. Points at
docs/ROADMAP.md and docs/README.md instead of the internal trackers.
No infra identifiers or internal mechanics remain; all links resolve.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two concrete, verifiable documentation gaps from the open-source review:
- docs/ had no index. Adds docs/README.md grouping the 60-odd published docs by
task — getting started, architecture, app development, design docs, ADRs,
security, roadmap — in the bitcoin/bitcoin doc/ style the plan called for.
Every link in it resolves (checked). The top-level README now points at it as
the front door rather than duplicating the list.
- ROADMAP.md and tests/lifecycle/TESTING.md linked docs/multinode-testing-plan.md,
which moved to local-only (it is a fleet node inventory, not published). Those
references now describe the scope split in prose instead of pointing at a file
that is not in the public tree.
The index is honest about what is missing: it names the three app-developer
docs the plan flagged as gaps (quadlet compilation, container lifecycle,
secrets materialisation) as not-yet-written, and points at the authoritative
code for each rather than pretending they exist.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The repo ships an MIT LICENSE and the README carries an MIT badge, but the
crates themselves declared no license, so `cargo metadata`, packaging and any
downstream mirror saw "license: null". Adds [workspace.package] license = "MIT"
and inherits it in all five members via license.workspace = true. Verified with
cargo metadata: all five now report MIT.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Today's outage window came from ordering, and the ordering was baked into the
publish script itself: it pushed main — the branch nodes read the manifest
from — together with the tag, up front, then uploaded and verified assets
afterward. So the manifest advertised the new version for the entire
upload+verify window. When an upload failed inside that window, every polling
node briefly saw a v1.7.126-alpha update whose binary 500'd and whose tarball
did not yet exist.
Reordered so the manifest goes live last:
1. push the TAG only (the Gitea release and asset URLs hang off it; the tag
alone changes nothing for nodes)
2. upload assets
3. verify every asset downloads in full and matches the manifest sha256/size
4. only then push main — the step that actually triggers nodes
Also fixes a way a bad asset could slip through unnoticed: the inline
verification ran in a `while read` pipe subshell, where its `fail` (exit 1)
terminated only the subshell and let the script continue to "published and
verified". Verification now runs in the main shell via a new
check-release-assets.sh, which fails hard on the first bad asset. The same
script is the reusable by-hand verifier used to recover today's release
(both assets confirmed 200 + sha256-match before the manifest was re-published).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
13-15's two artifacts. The map names plan/wave/threat-ref per row with
today's measured results (Rust assistant suite 130/130; adapter 37; broker
25; toolConfirm, audioPlayer 11, appsConfig 13; AIUI 353/356 with the three
documented pre-existing fixture failures). Four close-out rows added
(S-invariants, evals, egress, mock-free-bundle grep). Manual-only table
discharged except the physical-handset pass, which 13-UAT records as owed
(the AIUI-06 flagged assumption wants both, and only devtools-mobile is on
record). Open questions 1/2/4 resolved with plan cites; Routstr stays
honestly open.
Co-Authored-By: Claude <noreply@anthropic.com>
Restores the signed v1.7.126-alpha manifest to main now that both artifacts are
confirmed fetchable end-to-end:
- archipelago HTTP 200, sha256 matches the manifest
- frontend tarball HTTP 200, sha256 matches the manifest
The earlier publish was rolled back (e346e552) because the manifest went live
before its assets resolved. Two separate asset faults, now fixed: the binary's
first upload landed corrupt server-side and returned 500 on download (deleted
and re-uploaded, clean); the tarball's first upload returned an empty response
and never attached (re-uploaded, 201, full 210 MB).
This is byte-for-byte the manifest the tag already carries, so its signature is
unchanged and re-verified against the pinned release root.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The v1.7.126-alpha manifest went live on main — which is where nodes read it
from — before its artifacts were reachable. The binary returns HTTP 500 and the
frontend tarball never uploaded (404), so any node polling would advertise an
update it cannot fetch.
Restores the previously published, still-validly-signed .125 manifest
byte-for-byte from 19487670, so nodes see the last release that actually
resolves. The v1.7.126-alpha tag and its signed manifest are unchanged in git
history; only what main serves is rolled back.
Publishing order was the mistake: the manifest is the trigger, so assets must
be verified downloadable before it lands on main, not after.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed release manifest for v1.7.126-alpha, verified against the pinned
release root before committing.
Committed by hand rather than by re-running create-release.sh: the script
regenerates the manifest at step 6, which would overwrite the signature
applied at step 6b. Its own "sign it, then re-run this script" advice destroys
the thing it just asked for.
Version bump, changelog and What's New landed earlier in 1cd068e4 — the
frontend build embeds the version via the curated What's New list, so those
must exist before the build step rather than after it.
Release gate: 8 of 9 stages passed in-run. cargo-test-weekly hit its 1500s
ceiling (exit 124, a timeout not a failure) because the non-incremental
all-targets compile does not fit on this machine; the suites were then run
separately and passed 100/100, including the downgrade guard and 30 update::
tests over the OTA apply/rollback path. Right-sizing that ceiling is a
follow-up — an override that becomes routine stops being a gate.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Live on archi-dev-box: the node still proxied web-search straight to
SearXNG :8888 unauthenticated — the repo conf was fixed in d0c9ea6e but
existing nodes' /etc/nginx/sites-enabled never gets rewritten by a source
edit. Added to the nginx self-heal battery: stale 8888/search proxy_pass →
session-gated 5678 with the Cookie forwarded (heal_stale_web_search_block,
pure + idempotent + tested). Fresh ISOs already ship the gated block.
Co-Authored-By: Claude <noreply@anthropic.com>
Mesh AssistQuery answered with a bare LLM call — no tools, no actions.
The CallerScope::Mesh variant was designed for this wiring ('the variant
exists so the shape is right when a future plan wires mesh callers into
the shared loop'); this is that plan. A trusted/allowlisted asker's prompt
now runs assistant::chat with CallerScope::Mesh { authorized } — the
operator's persisted grants cap what the model may touch (never wider),
and writes suspend on the node's own confirm gate. The reply is capped
for airtime as before, with a brevity instruction for mesh turns.
Wiring follows the blob_store pattern: RpcHandler::set_mesh_service (now
&Arc<Self>) forward-propagates an Arc<RpcHandler> into the mesh state's
new assistant_handler slot; absent (early boot) falls back to the legacy
bare-LLM answer.
Test: mesh_caller_authority_is_capped_at_operator_grants.
Co-Authored-By: Claude <noreply@anthropic.com>
Build #189 proved the IPv4 fix works (downloads all succeeded) and proved
my own edit wrong: appending inet4_only to /etc/wgetrc before wget's
package landed made dpkg's conffile prompt hit EOF, leaving wget and
debootstrap unconfigured. Moved the pin below the apt install.
Co-Authored-By: Claude <noreply@anthropic.com>
The 2912×1632 jpg painted visibly slowly over Tailscale/Tor. 1920w q82
webp is visually identical behind glass and under half the weight.
Co-Authored-By: Claude <noreply@anthropic.com>
This box (and its containers) blackhole IPv6: deb.debian.org answers AAAA
first, wget tries v6 until debootstrap's per-package timeout, and the
installer-env stage died twice today with 'Couldn't download packages'.
Probed in a debian:trixie container: v4 OK, v6 hangs. inet4_only for wget
(covers debootstrap) + Acquire::ForceIPv4 for every apt-get, including the
chroot and the rootfs Dockerfile stages.
Co-Authored-By: Claude <noreply@anthropic.com>
Two changes, one binary batch:
1. app_install/app_uninstall (task 3): '!ai please install bitcoin knots'
correctly said it can't. Both tools are category-Apps, destructive, and
ride the 13-08 confirm gate (node-authored descriptions added). Install
validates catalog membership BEFORE the dialog (a typo never spends an
approval); uninstall resolves installed ids. Both reach the SAME
package.install/package.uninstall spawns every authenticated caller
uses, via a curated Arc-taking sibling of assistant_dispatch_tool.
2. S6: cloud legs no longer strip prior USER turns from replayed history.
Turn-minimality's allowlist is now the whole conversation's operator
turns (the node's own D-08 transcript, same trust class as this turn),
still mechanically matched, B1 secret scan and 64KB cap unchanged,
fabricated user messages still truncated. The model no longer sees its
own answers without the questions.
Co-Authored-By: Claude <noreply@anthropic.com>
resolveAppIcon's final arm guessed /assets/img/app-icons/<id>.png — strfry
404'd live. DEFAULT_APP_ICON already existed; the chain now ends on it.
Co-Authored-By: Claude <noreply@anthropic.com>
Node-derived cards have no rating/year/runtime/director; rendering the
defaults read as '★ 0 · 0m' beside an empty string. FilmGrid, FilmDetail
and SongGrid now gate those spans on real values (FilmCard, Book and Place
components already did). Panel empty states were made honest in 9abc1623
('Nothing found'), so this closes the rendering-honesty item.
Co-Authored-By: Claude <noreply@anthropic.com>
create-release.sh builds the frontend at step 4 and validates the curated
changelog at step 5, then requires the freshly built bundle to contain the new
version. The version reaches the bundle only through the hand-written What's
New list, so on a fresh release that check can only pass if the changelog and
What's New entries are written BEFORE the script runs. Writing them after is
what aborted the first attempt.
Leads with the downgrade bug, since that is the one users saw: an Update button
offering the release withdrawn for an actively exploited 2FA bypass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every mock consumer is now gated on the demo flag inline (canonical Vite DCE
idiom — the cross-module DEMO_CONTENT_ENABLED const defeated folding). But
the real leak was films.ts's module-level allGenres/allSources exports:
[...new Set(mockFilms.flatMap(...))] is unprovably pure, so the treeshaker
kept the whole module — array, plex:// and cloud.example.com hosts and all —
even with zero live references. The mocks directory is now declared
side-effect-free in vite.config (they are pure data by design), so unneeded
mock modules actually drop.
Verified: clean dist build → entry bundle AND dist-wide grep show zero
mock hosts (spotify/track/example, cloud.example.com, plex://, tmdb image
host). Demo/dev builds (VITE_DEMO_CONTENT=true or import.meta.env.DEV) keep
the full pack. Tests: 353/356, failures are the three documented
pre-existing ones.
Co-Authored-By: Claude <noreply@anthropic.com>
The release gate failed cargo-fmt. The domain that replaced the IP-based
registry is longer, pushing several test assertions past the width limit, so
rustfmt wanted to re-wrap them. Pure line re-wrapping — no semantic change.
Caught by the pre-flight gate rather than after tagging, which is what it is
for.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The legacy `btcpay` entry (distinct from `btcpay-server`, no embedded manifest)
still carried a concrete 2.3.9 image. catalog_primary_image treats that as
authoritative, which is what drove the UI to offer "update to 2.3.9" on nodes
already running 2.4.2 — a rollback onto the actively exploited release.
Now 2.4.2 in both entries, signed by the pinned release root and verified.
This lands the fix for every node immediately, without waiting for the binary
carrying the downgrade guard (cbfda305) to reach them. The guard remains the
durable fix: it makes any future stale pin fail safe rather than relying on
every version claim being correct.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
package.versions answered installedVersion "15.17" for btcpay-server while
offering "2.4.2" — 15.17 being its postgres dependency's tag. With BTCPay's
own container absent, installed_version fell back to `containers.first()`,
which for a multi-container stack is an arbitrary sibling.
That is the number the update decision is made from, and it is what the UI
shows next to the available version, so a nonsense pair like "installed 15.17,
available 2.4.2" is presented as a legitimate upgrade.
The fallback now only applies when there is exactly one container, which still
covers apps whose container is named differently from their id (immich_server
for immich). With several containers and no identifiable backend, the honest
answer is "unknown" rather than a guess at a sibling.
Extracted as select_backend_container so the rule is testable directly.
Tests: the BTCPay stack case, the lone differently-named container, and the
archy- prefixed preference. Full suite 1157/1157.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The app store offered "update to 2.3.9" on a node already running 2.4.2 — the
release that fixes an actively exploited 2FA bypass. Taking it would have
rolled the node back onto the vulnerable version.
Root cause: available_update_for_images compared tags for inequality only.
Same repo + different tag meant "update available", with no ordering. Every
version claim upstream of it can go stale — the signed catalog, a legacy
catalog entry, the image-versions.sh baseline pin — and any one of them
lagging turned into a backwards Update button.
Guard added: when both tags parse as dotted-numeric versions, a lower pinned
version is never offered. Tags that cannot be ordered (RELEASE.2024-11-07…,
14-vectorchord0.4.3) keep the previous behaviour rather than silently losing
updates. This makes stale data fail safe, which matters more than any single
pin being correct.
Four sources still named 2.3.9, three of them able to act on it:
- releases/app-catalog.json — a LEGACY `btcpay` entry, distinct from
`btcpay-server`, carrying a concrete 2.3.9 image. catalog_primary_image
treats that as authoritative, so this is what drove the button. Fixed, but
held back from this commit: it needs re-signing.
- scripts/image-versions.sh — the baseline pin used when the catalog does not
cover an app.
- stacks.rs — the legacy BTCPay installer, twice. The fallback install path
would have deployed 2.3.9 outright.
- neode-ui curatedApps/marketplaceData and public/catalog.json — the store's
displayed version, hardcoded rather than read from the catalog, which is why
it still showed 2.3.9 after the update landed.
Audited every other installer for the same shape. The remaining literals are
the immich stack, which currently agrees with its manifests; hits in
set_config.rs and app_catalog.rs are test fixtures. To keep it that way,
scripts/check-installer-image-pins.py asserts that any installer literal
naming the same repository as an app manifest carries the same tag, and runs
blocking in CI. Verified it catches a simulated revert to 2.3.9.
Tests: 13/13 in image_versions including the exact BTCPay case, a genuine
upgrade still offered, equal versions silent, prerelease suffixes ordered on
their numbers, and opaque tags unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
d0af38e8 shipped check-catalog-registry-trust.py without the file it reads:
`releases/**` in .gitignore silently dropped registry-trust-floor.json, so the
guard would have failed in CI and on any fresh clone. app-catalog.json only
stays tracked because it predates that rule.
Both are source rather than build output — nodes fetch the catalog from this
path on main, and the floor is what the guard checks it against — so both now
have explicit negations, with the reason recorded next to them.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Encodes the sequencing rule that nearly shipped a fleet-wide outage today.
The signed catalog is authoritative on every node — catalog_image_override
makes its image refs beat the on-disk manifest. TRUSTED_REGISTRIES in the
working tree describes a binary being built now; nodes run whatever was last
shipped to them. Those two diverge for exactly as long as an OTA takes to
reach the fleet, and that window is when regenerating the catalog silently
breaks every install with "not from a trusted registry".
Regenerating today would have done precisely that: the generator embeds each
app's manifest, and those now name the new registry domain, which no deployed
binary trusts.
- releases/registry-trust-floor.json records the hosts DEPLOYED binaries
trust, separately from what the source tree accepts, with the new domain
parked under `pending` until an OTA carries it. The migration order is
written down there rather than living in someone's memory.
- scripts/check-catalog-registry-trust.py compares the catalog's hosts against
that floor and explains the ordering fix when they diverge.
- sign-catalog.sh runs it as a preflight BEFORE prompting for the mnemonic, so
a bad catalog is refused at the last reversible moment.
- CI runs it blocking, plus the drift report advisory (drift between a manifest
landing and the next signed release is expected, since only the ceremony can
close it).
Also installs PyYAML in the manifests job. That job passed only because GitHub
runners happen to ship ruby, which the validator used to require; it now needs
python3+PyYAML.
Verified: passes on the published catalog (2 hosts, both trusted); refuses a
simulated full regenerate (79 refs on the untrusted domain) and blocks the
ceremony without requesting the mnemonic.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed by the pinned release root and verified before publishing
(`ceremony verify` → OK). A present-but-invalid signature is a hard reject on
nodes, so verification is the gate, not the presence of a signature field.
Surgical edit rather than a regenerate: only the btcpay-server entry changed
(66 apps in, 66 out; 49 entries still resolve through the legacy registry,
untouched). A full regeneration would have embedded the repo's manifests,
which now name the new registry domain that no deployed binary trusts yet —
publishing that would have broken every app install fleet-wide, during a
security push. That sequencing is being fixed separately.
BTCPay's image comes from docker.io, so it is unaffected by the registry
migration either way.
Nodes pick this up via package.check_updates → refresh_catalog → reload
manifests, then package.update (stop → pull → remove → recreate → verify,
with rollback on failure).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
BTCPay 2.4.2 fixes a critical vulnerability that upstream reports as actively
exploited: a TOTP two-factor bypass via Greenfield Basic authentication
(btcpayserver/btcpayserver#7491).
Checked the 2.3.9 -> 2.4.2 breaking changes against how Archipelago actually
configures BTCPay; both are inert here:
- 2.4.0 removed the LNBank and Lightning Charge backends. Ours is a direct LND
connection built by container::lnd::ensure_btcpay_lnd_connection_secret.
- 2.4.2 disables Greenfield Basic auth five minutes after account creation.
Nothing in the daemon or frontend consumes BTCPay's API.
The manifest bump alone does NOT reach nodes: catalog_image_override makes the
signed catalog authoritative whenever the image repo matches, so a node would
be forced back to 2.3.9. The catalog edit is held locally until the signing
ceremony runs, because an unsigned catalog published to main would be accepted
by nodes (absent signatures are allowed) and would quietly drop authenticity.
check-app-catalog-drift.py only understood app-catalog/catalog.json, where
`apps` is a list. releases/app-catalog.json — the SIGNED catalog nodes actually
resolve apps through — keys `apps` by id and wraps each app's full manifest
under manifest.app. So the checker parsed the file that governs nothing and
raised ValueError on the file that governs everything. It now reads both shapes.
Running it against the release catalog shows the repo and the catalog agree on
content: of 34 image differences, all 34 are the registry host alone and every
tag is identical. The remaining version-string drift (v1.18.0 vs 1.18.0,
1.30.0-alpine vs 1.30.0) is cosmetic metadata, not image drift.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Fixes 4 test failures introduced by 8e814ca0, which I pushed after running
only the container-crate tests while the full suite was still compiling. Both
failures were real defects, not stale assertions.
1. Catalog-driven installs would have failed fleet-wide.
8e814ca0 dropped the old registry address from TRUSTED_REGISTRIES, but the
signed catalog still advertises image refs on it — deliberately, since
rewriting a signed artifact invalidates its signature. Nodes resolve apps
through the catalog, so every install would have been refused with "not
from a trusted registry". Reinstated as LEGACY_REGISTRY_HOST, documented
as transitional and removable only once the catalog is re-signed.
2. The update fallback lost the property it exists for.
update.rs keeps two mirrors on purpose: the domain as primary, and the
old IP over plain HTTP as a fallback, because a node whose DNS or clock is
wrong (both break TLS) must still be able to update itself — the signature,
not the transport, is what makes either source safe. The bulk rewrite
pointed both constants at the domain, leaving the escape hatch dependent on
exactly what it exists to survive. Restored to its original value.
Separately, validate-app-manifest.sh is ported from ruby to python3+PyYAML.
It shelled out to ruby with stderr discarded, so on any machine without ruby
a missing interpreter was reported as "Valid YAML with top-level app block:
FAIL" and every manifest came back REJECTED. This is the first tool an app
developer runs, and it sent them to fix YAML that was never broken. Ruby was
also the odd dependency out — the repo already ships three python scripts.
It now checks for python3 and PyYAML up front and names what is missing, then
parses with PyYAML. Missing keys resolve to an absent-value object that
indexes to itself and prints empty, so call sites lost their per-hop guards:
(((app["container"] || {})["build"] || {})["context"])
becomes app["container"]["build"]["context"]. Booleans still print as
true/false rather than Python's True/False — call sites compare == "true",
so Python's capitalisation would have silently inverted the readonly_root
and no_new_privileges security checks.
Verified: full rust suite 1148/1148, 0 failed. All 56 app manifests validate
(0 rejected, 0 errored) where previously every one was rejected. No signed
artifact modified.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
RC1's aiui/index.html pointed at the stale checked-in demo bundle while
today's dist sat beside it unreferenced: demo/aiui was tried first, and the
rsync without --delete merged it over the fresh capture from /opt. Now the
newest index.html across all candidates wins (demo/aiui remains the
fresh-clone fallback) and the copy deletes before writing.
Found by mounting the RC1 ISO and diffing bundle hashes against the tree —
exactly the 'verify the frontend INSIDE the ISO' rule.
Co-Authored-By: Claude <noreply@anthropic.com>
Live evidence, two ways: the 9abc1623 banner never fired because D-16 hides
ungranted tools (model never calls → refused_categories always empty), and
the [[needs:id>]] marker fix failed because a small local model answers with
a workaround narrative instead of emitting structured markers.
The model's reliable, trained behavior is tool CALLING — so disabled tools
are now listed in a DISABLED prompt section and remain in the schema. A call
hits the execution gate, which refuses and records the category → the
trusted chrome offers Settings → AI Data Access. Deterministic and
model-independent. The prompt split is UX/attack-surface shaping; the
security boundary remains the server-side grant re-check in execute_tool
(loop_.rs), unchanged and now the single enforcement layer by design.
Tests: ungranted_tool_only_ever_in_disabled_section (section-aware),
disabled_tools_are_listed_as_callable_but_refused, marker extraction kept
as a harmless safety net. 127/127 assistant suite green.
Co-Authored-By: Claude <noreply@anthropic.com>
Replaces the registry host across 86 files: 309 references, covering all 40
app manifests, the orchestrator and container crates, the release and catalog
scripts, both demo-images workflows, the ISO builder, demo-deploy, and the
frontend marketplace data.
Verified the domain actually serves the registry before rewriting anything,
rather than assuming the web host implies the registry:
- TLS verifies clean, HTTP/2 on the web root
- an anonymous token grants a manifest fetch (HTTP 200) with no credentials
- skopeo inspect --no-creds resolves an image and lists its tags
That last check is the one that matters: an outside developer with no account
can now pull, which was the functional blocker for publishing at all.
Plain-HTTP references become HTTPS in the same pass, so OTA downloads stop
crossing the network in the clear.
Deliberately NOT rewritten:
- The public FIPS anchor on port 8444. It is a functional network endpoint
every node dials to bootstrap the mesh — closer to Bitcoin Core's hardcoded
seeds than to leaked infrastructure. The domain does resolve to the same
host, so it could become a hostname, but that adds a DNS dependency to the
path used precisely when things are broken. Worth a deliberate decision,
not a side effect of this change.
- The companion APK on port 2100. The domain returns 404 for that path, so
rewriting it would swap a working URL for a broken one. The Releases page
does serve (200), which is where the plan already wants those binaries.
- releases/app-catalog.json, releases/manifest.json and release-manifest.json.
These carry `signature` and `signed_by`; editing their contents invalidates
the signature and the fleet refuses artifacts that fail verification. They
were rewritten in a first pass and reverted — they must be regenerated and
re-signed through the signing ceremony instead, which needs the mnemonic.
So the catalog still advertises the old host until that ceremony runs. Nodes
resolve images through the signed catalog, not the on-disk manifests, so this
commit alone does not change what a node pulls.
Verified: archipelago-container 75/75; every manifest still parses with a
top-level app block; no signed artifact modified.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The 9abc1623 banner waited on refused_categories, but refused_categories
only fills when the model CALLS a gated tool — and D-16 hides ungranted
tools from the prompt, so the model never calls: it answered 'I can't do
that' in prose and the banner never fired. Live-verified: revoke media,
ask for content, no banner.
- build_system_prompt takes the disabled categories and teaches a marker:
'say it can be switched on in Settings → AI Data Access and end with
[[needs:<id>]]' — category names only, never tool names (D-16 holds)
- extract_needs_markers strips the markers from the reply and folds them
into refused_categories; unknown ids pass through as text (an offer is
the worst a bad marker can cause — never a grant)
- egress's seed-screen test now covers the new paragraph too
Co-Authored-By: Claude <noreply@anthropic.com>
Keeps the dev and test tooling an outside contributor would want, and takes
our node addresses out of it.
Scripts that silently defaulted to one of our nodes now require an explicit
host and exit 2 without one: smoke-test.sh, trust-archipelago-cert.sh,
dev-container-test.sh (which also derives its RPC and health URLs from the
SSH target instead of a second hardcoded copy), and image-recipe/dev-branding.sh.
A default that points at a machine the user does not own is worse than no
default: it fails confusingly, or reaches a stranger's device.
Usage examples, mock data and test fixtures move to the RFC 5737
documentation range (192.0.2.0/24). CGNAT test values stay inside
100.64.0.0/10 so the range-check semantics they exercise still hold, and
192.168.1.0/.1/.254 are left alone — those are gateway logic and UI
placeholders, not our addresses.
Playwright and the perf spec defaulted their baseURL to one of our nodes;
they now default to localhost:8100, the local dev server.
Removed neode-ui APP_URLS entirely. It is dead code — exported, never
imported — and it pinned fedimint's *prod* launch URL to 192.168.1.228:8175.
Had anything consumed it, every user's node would have tried to reach an
address that on their LAN is either nothing or someone else's machine.
Deleting beats sanitizing dead config.
Verified: frontend 868/868 vitest across 108 files; archipelago-container
75/75; mesh tests 9/9; audit-secrets 5/5. Zero node addresses and zero node
names remain in tracked files.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The repo is source code and guidelines only. Nothing about how Archipelago's
own fleet is run, or how the team works, stays in it.
Untracked (kept on disk, gitignored) — 250 files:
- .planning/ (199) and loop/ — internal development process
- fleet operations tooling that targets specific nodes: deploy-to-target,
deploy-tailscale, deploy-config-defaults, setup-target-dev, setup-aiui-server,
setup-https-dev, debug-frontend, node-profile, fleet-fips-pair/unpair,
image-recipe/sync-from-live.sh
- image-recipe/INTEGRATION-GUIDE.md and docs/multinode-testing-plan.md, both of
which are live-server workflow and fleet node inventories
- the Phase 10 on-node verification and evidence records, which cite .planning/
as their evidence base
KEY-05-ENTROPY-ENFORCEMENT.md was initially moved out with the other Phase 10
docs and then put back: it is cited as normative rationale from ten places in
the codebase, including core/clippy.toml, which bans rand::thread_rng and
points at it for the reason. That makes it a guideline, not an internal record.
Node names removed from source (48 occurrences across comments, manifests and
test fixtures): archi-dev-box, archy-x250*, shorty-s, framework-pt,
zaza-optiplex, archi-thinkpad. Comments keep the engineering context and the
date, which is what carried the meaning; the machine name did not.
Three of those were live test values rather than comments and were replaced
with valid stand-ins, not prose: two mDNS hostnames and a mesh peer name.
An earlier pass substituted "a test node" into a hostname assertion, producing
an invalid hostname; caught and fixed as test-node.local.
Wipe mechanism: .local-only/manifest.txt inventories every local-only path and
.local-only/wipe.sh deletes them on one confirmation, refusing to touch
anything git still tracks. Both are themselves untracked, so the public repo
does not carry a map of internal filenames.
Verified: cargo check -p archipelago --all-features clean; archipelago-container
75/75 tests pass; appOrigin vitest 7/7; audit-secrets 5/5; every relative link
in tracked markdown resolves (0 broken).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
usePlayer.play() never looked at song.sources[] — a real library track went
straight to (CSP-blocked) Wavlake and reported 'Not found on Wavlake' while
its bytes sat on the operator's disk. Node sources (same-origin /content/<id>,
Range-streamed) now play first; Wavlake is the metadata-only fallback.
FilmDetail likewise only played YouTube sources; own/peer/IndeeHub sources
(same-origin, media-src 'self') now win, YouTube stays the free-films
fallback.
Co-Authored-By: Claude <noreply@anthropic.com>
Untracking the ops docs broke every reference to them. Repoints or removes
those references across README, architecture, ROADMAP, the archive index,
lifecycle TESTING, the hardening plan, and the security docs — pointing at
the issue tracker where a live task list was meant, and dropping the entry
entirely where it only existed to link an internal file.
Also fixes two pre-existing broken links found by validating every relative
link in the tracked docs:
- README linked docs/OPEN_SOURCE_READINESS.md, which never existed
(underscores vs hyphens).
- reticulum-daemon/README.md linked a local Claude session plan at
../../.claude/plans/enchanted-strolling-rocket.md — outside the repo, and
a path that would have shipped publicly pointing at nothing.
All relative links in tracked markdown now resolve: 0 broken.
audit-secrets.sh still 5/5.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Operations docs move out of git entirely rather than being sanitized. They
stay on disk for local use and are gitignored, so the Phase 6 export (which
takes HEAD) can never carry them. 15 files: the fleet runbook, hotfix
process, node inventories, internal trackers, session handoffs, the key
rotation/signing-posture records, and the open-source plan itself.
For the docs that remain public, infra identifiers are replaced with things
that are better documentation rather than placeholders: curl examples now
use `archipelago.local`, the product's own mDNS name, so a reader can run
them as-is instead of substituting an address that was never theirs.
Deliberately NOT scrubbed, both verified as functional rather than leaked:
- `tx1138.com` is the shipped default block explorer (DEFAULT_TX_EXPLORER in
useTxExplorer.ts, surfaced in WalletSettingsModal). Product behavior.
- `git.tx1138.com` in core/container/{image_policy,registry}.rs is a retired-
registry constant the code matches on to strip stale entries from legacy
node configs. Removing it would break migration for older nodes.
- `192.168.1.254` in bulletproof-containers.md is the LAN gateway in a podman
bug description, and `192.168.1.x` in user-walkthrough.md is already generic.
Whether a personal domain should be the shipped explorer default in a public
product is a separate product question, not a security one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Sequencing change per user decision: credential rotation/revocation runs
last, immediately before Phase 6, instead of first. Safe under fresh-history
publish — the scrub commits never become public — but recorded as a HARD
blocking gate on Phase 6, with an explicit rotation sign-off added as a
numbered pre-publish step so "scrubbed" cannot be mistaken for "rotated".
Also corrects the plan against what execution actually found:
- the fleet password was in 8 tracked files, not 7 (3 in .planning/)
- both Gitea tokens are already dead (401); only the `ai` password is live
- the Framework node's SSH password was rotated out-of-band and is unrecorded,
which would block it from receiving the fleet rotation
- .planning/ is 199 tracked files of internal agent state — added to Phase 2
as the largest un-triaged internal block still in the tree
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Scrubs the fleet SSH/UI password from every tracked file (22 occurrences)
and removes inline credentials from the code paths that used them.
Docs and trackers keep the surrounding context — these are published under
docs/history/ per the open-source plan — with the literals replaced by
<FLEET_PW> / <FLEET_PW_ALT> so the "two variants exist" detail survives
without the values.
Three of the eight files were in .planning/ and were NOT in the plan's
enumerated list; the reworked audit-secrets.sh found them.
Code changes:
- neode-ui/test-openwrt.mjs: node URL and password come from ARCHY_NODE_URL /
ARCHY_NODE_PW; the SSH target derives from the URL instead of a hardcoded
tailnet IP; exits 2 when unset.
- scripts/run-post-install-tests.sh: drops the built-in "testpass123!"
default and adds --password-stdin; refuses to run unauthenticated instead
of silently trying a known password. --phase1-only still needs no password.
- .gitea/workflows/post-install-tests.yml: sshpass with an inline literal
replaced by key auth (NODE_SSH_KEY secret); password comes from the
NODE_UI_PASSWORD secret and is piped over stdin rather than argv, so it
stays out of the node's process list and the job log. Default target IP
removed.
scripts/audit-secrets.sh now reports 5/5 pass, 0 fail.
Note: rotation of the exposed credentials is deliberately deferred to the
pre-publish gate and is NOT done by this commit — these values are still
live. See Phase 0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
