Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Bumps [markdown-it](https://github.com/markdown-it/markdown-it) from 14.1.0 to 14.3.2.
- [Changelog](https://github.com/markdown-it/markdown-it/blob/14.3.2/CHANGELOG.md)
- [Commits](https://github.com/markdown-it/markdown-it/compare/14.1.0...14.3.2)

---
updated-dependencies:
- dependency-name: markdown-it
  dependency-version: 14.3.2
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>

WebsocketNetwork.Stop has the same gap. innerStop sweeps wn.peers
before the HTTP server shuts down, so a handshake that completes in
between is registered. Nothing closes that peer. Shutdown ignores
hijacked connections and the wsPeer loops don't watch wn.ctx, so the
peer outlives Stop.

innerStop now sets peersClosed under peersLock, and addPeer rejects
late peers. addPeer reports whether the peer is registered. ServeHTTP
and tryConnect close a rejected peer and skip the connected logs and
telemetry, which also applies to the existing didSignalClose rejection.

P2PNetwork.Stop sweeps wsPeers in innerStop before it closes the host.
A stream handler that completes in between registers a wsPeer that
nothing closes. Once host.Close resets its stream, the peer's read loop
logs and emits telemetry, possibly after Stop returns.

innerStop now sets wsPeersClosed under wsPeersLock, and
baseWsStreamHandler checks it in the same critical section as the
insert. A rejected peer releases its identity and throttle slot and is
closed.

Found while reviewing #6744.

* Show ARC-56 signing risk icon next to Sign all, using registry owner reputation

Adds an at-a-glance icon (trusted / warning / danger / not-ABI) beside the
"Sign all" button in WalletConnect/Liquid requests and on /signAll, driven by
the registry trust level plus the new per-owner reputationScore/riskLevel/banned
fields from the ARC56Registry owners files. Publishers now show their rating.
Decode logic is extracted into a shared composable; risk evaluation is a pure,
unit-tested module.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Address review: harden risk verdict, dedupe decode, accessible popover

- Non-ABI verdict no longer neutral (warning colour, 'not verifiable')
- App create/update/delete and close-out/rekey in the request block a trusted verdict
- Share in-flight decode between summary and icon; registry fetch timeout
- Click/keyboard popover instead of hover-only tooltip
- Reword trusted text (acceptable reputation), fix doc comment placement

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Address second review: registry timeout, no caching of transient failures, stricter trust

- Actually apply registry fetch timeout; do not cache network-failure misses
- Only low-risk publishers are trusted; ClearState counts as sensitive
- Risky fields (close-out/rekey) raise a caution even without app calls
- Every app call gets a risk input (keeps sensitive flag when decode fails)
- Safe in-flight dedupe key, owner badge computed once, aria-haspopup/expanded

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Address third review: non-ABI ban check, more risky txn types, verdict deadline

- Hash non-ABI calls so a banned publisher is still caught
- Clawback/freeze/keyreg/rekey-away/close-out raise a caution; ignore self/zero rekey
- Only network failures (not non-JSON mirrors) skip negative caching; 5s fetch timeout
- 20s overall verdict deadline (fail-closed), short-lived shared decode results
- Clear stale summaries on new decode, close popover on reload, tidy owner links
- Fix changelog wording

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Address fourth review: lookup-failed reason, single decode map, tested txn rules

- Failed/timed-out lookups get their own reason and are never cached
- Non-ABI with no lookup fails closed; banned non-ABI is danger
- Pure, unit-tested isRiskyTransaction/isSensitiveAppCall (acfg reconfigure, CloseOut added)
- One shared-decode map; hide 'no publisher' noise for plain non-ABI calls
- Trusted text scoped to contract calls; owner badge colours match verdict

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Address fifth review: distinguish registry outage from not-found

- Strict registry lookups throw RegistryUnavailableError (owners, primary spec); 5xx/429 no longer cached as 404
- algod failure for an existing app surfaces as lookup_failed; creation uses the txn program
- Identity-checked cache eviction, node in shared-decode key, single owners computed

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Address sixth review: mirror-aware outage detection, ban check on every branch

- A definitive 404 from any mirror beats another mirror's outage; truncated bodies are transient
- Banned publisher is danger for unknown/selector-only/verified/non-ABI alike
- Arc56CallDetails resolves programs like the shared decode (creation handled)
- Locale reason texts cover every trigger; type-only Arc56TrustLevel import

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Ludovit Scholtz <ludovit.scholtz@aaaauto.cz>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>

- A definitive 404 from any mirror beats another mirror's outage; truncated bodies are transient
- Banned publisher is danger for unknown/selector-only/verified/non-ABI alike
- Arc56CallDetails resolves programs like the shared decode (creation handled)
- Locale reason texts cover every trigger; type-only Arc56TrustLevel import

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

- Strict registry lookups throw RegistryUnavailableError (owners, primary spec); 5xx/429 no longer cached as 404
- algod failure for an existing app surfaces as lookup_failed; creation uses the txn program
- Identity-checked cache eviction, node in shared-decode key, single owners computed

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

- Failed/timed-out lookups get their own reason and are never cached
- Non-ABI with no lookup fails closed; banned non-ABI is danger
- Pure, unit-tested isRiskyTransaction/isSensitiveAppCall (acfg reconfigure, CloseOut added)
- One shared-decode map; hide 'no publisher' noise for plain non-ABI calls
- Trusted text scoped to contract calls; owner badge colours match verdict

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

- Hash non-ABI calls so a banned publisher is still caught
- Clawback/freeze/keyreg/rekey-away/close-out raise a caution; ignore self/zero rekey
- Only network failures (not non-JSON mirrors) skip negative caching; 5s fetch timeout
- 20s overall verdict deadline (fail-closed), short-lived shared decode results
- Clear stale summaries on new decode, close popover on reload, tidy owner links
- Fix changelog wording

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

- Actually apply registry fetch timeout; do not cache network-failure misses
- Only low-risk publishers are trusted; ClearState counts as sensitive
- Risky fields (close-out/rekey) raise a caution even without app calls
- Every app call gets a risk input (keeps sensitive flag when decode fails)
- Safe in-flight dedupe key, owner badge computed once, aria-haspopup/expanded

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

- Non-ABI verdict no longer neutral (warning colour, 'not verifiable')
- App create/update/delete and close-out/rekey in the request block a trusted verdict
- Share in-flight decode between summary and icon; registry fetch timeout
- Click/keyboard popover instead of hover-only tooltip
- Reword trusted text (acceptable reputation), fix doc comment placement

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Adds an at-a-glance icon (trusted / warning / danger / not-ABI) beside the
"Sign all" button in WalletConnect/Liquid requests and on /signAll, driven by
the registry trust level plus the new per-owner reputationScore/riskLevel/banned
fields from the ARC56Registry owners files. Publishers now show their rating.
Decode logic is extracted into a shared composable; risk evaluation is a pure,
unit-tested module.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>