fix: round-2 full-repo architecture audit (21 waves + shielded-screen consolidation)

CHANGELOG.md

@@ -59,6 +59,16 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
   separate "fund directly from a specific transaction output" path was removed so
   there is a single, double-spend-safe funding flow.
 
+- **DashPay profile no longer requires a display name**: you can save your profile
+  with the display name left blank; only the length limits on name, bio, and
+  avatar URL are still enforced.
+
+- **Shield, Send Privately, and Unshield are now one screen**: each of these
+  actions opens the same Send screen, already set up for what you clicked, instead
+  of a separate dedicated screen per action. The steps and options are unchanged,
+  including pasting a raw shielded address in hex form — there are just fewer
+  screens to navigate between.
+
 ### Known Limitations
 
 - **Single-key wallets — send and balance refresh not available**: importing a
@@ -148,3 +158,14 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
   stay on screen until you dismiss them yourself. Routine notices (a
   successful action, a validation hint, an in-progress status) still clear
   automatically as before.
+- Choosing a funding method on the identity registration or top-up screen and
+  then switching wallets no longer silently discards your choice and reverts
+  to the default funding method.
+- The top-up screen's automatic wallet selection no longer pre-selects a
+  wallet whose spendable balance is too small or currently locked to actually
+  cover the top-up, which would then be immediately rejected.
+- The "My Tokens" tab no longer gets stuck on a loading spinner when you have
+  no identities yet; it now shows the expected empty state.
+- Two settings changes made in quick succession could occasionally cause one
+  of them to be silently lost. Saving settings is now a single atomic step,
+  so no change is dropped.

CLAUDE.md

@@ -271,7 +271,7 @@ Screens hold `Arc<AppContext>` and manage their own UI state.
 
 ### ConnectionStatus (single source of truth for connection health)
 
-`ConnectionStatus` (`src/context/connection_status.rs`) is the **single source of truth** for all high-level connection health state — RPC, ZMQ, SPV, and DAPI. For connection health (status, peer counts, errors, overall state), always read from `ConnectionStatus`, not directly from `SpvManager` or other subsystems.
+`ConnectionStatus` (`src/context/connection_status.rs`) is the **single source of truth** for all high-level connection health state — SPV and DAPI. For connection health (status, peer counts, errors, overall state), always read from `ConnectionStatus`, not directly from `SpvManager` or other subsystems.
 
 SPV status is **push-based**: `SpvManager` event handlers write directly to `ConnectionStatus` atomics (status, peer count, errors) as events arrive. The UI frame loop calls `refresh_state()` to recompute `overall_state` from these atomics, but does not poll SPV for health. This means `ConnectionStatus` is up-to-date in both GUI and headless/test contexts. Detailed SPV sync progress (heights, phase summaries used by tooltips) may still be read directly from `SpvManager.status()` until that progress reporting is migrated into `ConnectionStatus`.
 
@@ -327,4 +327,4 @@ Single SQLite connection wrapped in `Mutex<Connection>`. Schema initialized in `
 
 Linux (x86_64/aarch64), Windows (x86_64), macOS (x86_64/aarch64 with code signing)
 
-Requires protoc v25.2+ for protocol buffer compilation. Different ZMQ libraries for Windows (`zeromq`) vs Unix (`zmq`).
+Requires protoc v25.2+ for protocol buffer compilation.

Cargo.toml

@@ -105,19 +105,12 @@ native-dialog = "0.9.0"
 raw-cpuid = "11.5.0"
 
 [features]
-default = ["identity-hub"]
+default = []
 testing = []
 bench = []
 mcp = ["dep:rmcp", "rmcp/server", "rmcp/macros", "rmcp/transport-streamable-http-server", "dep:axum", "dep:subtle"]
 cli = ["dep:rmcp", "rmcp/server", "rmcp/macros", "rmcp/client", "rmcp/transport-io", "rmcp/transport-streamable-http-client-reqwest", "dep:clap", "dep:clap_complete"]
 headless = ["cli", "mcp"]
-# New unified Identities hub UI section (four-tab: Home/Contacts/Activity/Settings).
-# Default-enabled. Disable to get a smaller compile surface when iterating on legacy
-# Identities / Dashpay screens only.
-identity-hub = []
-# Unified Activity timeline aggregator. Off by default — the aggregator backend does
-# not exist yet; when off, the Activity tab renders a gated "coming soon" message.
-identity-hub-activity-feed = ["identity-hub"]
 
 [dev-dependencies]
 egui_kittest = { version = "0.35.0", features = ["eframe"] }
@@ -155,7 +148,7 @@ debug = "line-tables-only"
 
 [lints.rust.unexpected_cfgs]
 level = "warn"
-check-cfg = ["cfg(tokio_unstable)", "cfg(feature, values(\"testing\", \"bench\", \"mcp\", \"cli\", \"headless\", \"identity-hub\", \"identity-hub-activity-feed\"))"]
+check-cfg = ["cfg(tokio_unstable)", "cfg(feature, values(\"testing\", \"bench\", \"mcp\", \"cli\", \"headless\"))"]
 
 
 [lints.clippy]

docs/ai-design/2026-07-08-secret-decrypt-dedup/README.md

@@ -0,0 +1,58 @@
+# Secret-decrypt dedup (audit R2, Wave 1 — CODE-003)
+
+**Date:** 2026-07-08
+**Scope:** `model/wallet/encryption.rs`, `wallet_backend/{secret_access,single_key_entry}.rs`
+
+## Decision: local dedup (route a), NOT the upstream per-secret migration (route b)
+
+CODE-003 flagged the AES-256-GCM decrypt sequence (derive Argon2 key → init cipher
+→ checked-nonce → decrypt) as copy-pasted across three readers:
+
+- `wallet_backend/secret_access.rs::decrypt_hd_seed` — HD-seed migration reader
+- `wallet_backend/single_key_entry.rs::SingleKeyEntry::decrypt` — imported single-key reader
+- `model/wallet/encryption.rs::ClosedKeyItem::decrypt_seed` — deprecated `src/database` seed store
+
+The triage rationale proposed adopting the upstream platform per-secret encryption
+(XChaCha20-Poly1305 Tier-2 via `SecretStore::set_secret`/`get_secret`, already wired
+in `secret_seam.rs` as `put_secret_protected`/`get_secret_protected`) and deleting the
+local code.
+
+**We did NOT do that, deliberately.** Those three AES-GCM decrypts are the **legacy
+migration reader** for wallet secrets already written to users' disks in DET's own
+AES-GCM envelope format. Two authoritative module docs establish this:
+
+- `secret_seam.rs` module doc: *"The only remaining legacy decrypt (for not-yet-migrated
+  AES-GCM secrets) lives in the legacy reader."*
+- `wallet_seed_store.rs` module doc: the `envelope.v1` row *"is retained DECODE-ONLY as a
+  migration reader … rewritten to the raw label on the first load/unlock and then deleted."*
+
+The Tier-2 primitive already exists and the lazy legacy→Tier-2 re-wrap already runs in
+`secret_access::decrypt_jit`. Route (b) is therefore a **data migration** whose write side
+is already built — the readers must stay until every user's on-disk secret has been
+re-wrapped. Deleting them would lock users out of existing wallets.
+
+## What was done (route a)
+
+Extracted one crate-private `decrypt_message(ciphertext, salt, nonce, password, site)`
+in `model/wallet/encryption.rs`, returning `Zeroizing<Vec<u8>>` and a two-variant
+`DecryptError` (`WrongPassword` = AEAD auth failure; `Malformed` = structural/corrupt).
+All three readers route through it, mapping the two variants to their existing domain
+errors. Behavior is preserved exactly (same `TaskError` mapping for wrong-password vs
+corrupt-blob; same length validation at each caller). Structural diagnostics are logged
+once inside the helper with a `site` field.
+
+## Consequence for Wave 16 CODE-087
+
+**CODE-087 still applies, unchanged.** It targets the `(Vec<u8>, Vec<u8>, Vec<u8>)` triple
+returned by `encrypt_message` (+ its `type_complexity` allows) in
+`model/wallet/encryption.rs`. Route (a) keeps `encryption.rs` and `encrypt_message`
+exactly as-is, so the named-struct cleanup CODE-087 asks for is neither resolved nor
+mooted by this change — schedule it as originally planned.
+
+## If route (b) is ever scheduled
+
+It is a standalone migration task, not a dedup: extend the existing lazy re-wrap so the
+`ClosedKeyItem` (deprecated `src/database`) path is also migrated, confirm all three
+legacy formats have a re-wrap path, keep the readers until telemetry/versioning shows no
+un-migrated secrets remain, then delete the readers and `encryption.rs` together. Route (a)
+does not block route (b).