Use generic MCP result rendering and place labels above values on macOS and iOS. Update existing presentation expectations.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Remove low-value rendering, dependency-owned and script tests together with orphaned test scaffolding and obsolete wrappers. Keep state-driven visual coverage and release screenshot workflows.
Preserve incoming message origins on preparation failures, refresh tool presentation fixtures, and move shared SSH import coverage out of the iOS host test bundle to avoid duplicate static types.
Relocate the host icon input manifest, remove the heuristic AppStrings checker, and simplify localization CI.
Inject tool-owned presentation profiles and semantic vocabulary, remove legacy inference and stream compatibility, and update AgentKit to 3.1.0.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Tool names are protocol surface: they appear in the MCP export list,
persisted disabled-tool preferences, audit records, and presentation
registries. Spelling them as literals at each of those sites let one
place drift from another with no compile error to catch it.
Add TermindToolName as the single source of truth for the 25 names
Termind owns, and route every non-documentation reference through it.
The registry lives in TermindApplication because presentation cannot
import TermindRemoteInfrastructure, and TermindApplication cannot depend
on it either.
Names owned by AgentKit (request_user_secret, present_plan, the
scratch_* build tools) and retired aliases (ssh_exec, ssh_wait) stay as
literals, as do the prose in MCP instructions and descriptor summaries.
The legacyToolNames compatibility list is a historical data contract and
is unchanged.
This also removes the sshExecRunToolName and sshExecStdinToolName
constants from TermindAgentPolicy, which the registry replaces.
Track ET data-route progress separately from SSH bootstrap, preserve established hop colors, and reject stale route events. Distinguish direct Mosh recovery from full SSH reconnection while retaining whole-route failure coloring.
Consolidate Journey waiting transitions and failure presentation, reduce redundant rendering, and add route lifecycle and platform regressions.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Separate interaction refreshes from layout invalidation on macOS and iOS. Retain active editors across edit-eligibility changes and cover hosting identity, input composition, navigation controls, and model-driven streaming replay.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Share default transport ordering and connection-kind resolution across desktop and mobile opening paths. Keep explicit transport choices intact and skip unavailable default terminals before admission.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Centralize Journey mapping, interaction ordering, and localized tool presentation; remove redundant multiplexer and paywall state. Consume AgentKit 2.2.0 for shared operation, output, and transcript contracts, and correct its maintenance checkout documentation.
Validation: AgentKit 304 tests and local-package macOS/shared 215 regressions passed. iOS app built; 38 focused tests reported passing with an existing broken fixture excluded, but Xcode finalization stalled. Default macOS tests retain timeouts and Secure Enclave failures. A subsequent remote-pin build reported AgentTranscriptMessage.toolResult unavailable; it has not been diagnosed.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Integrate AgentKit 2.0.1 with shared scratch storage, whole-line tool policies, generated capture paths, and external MCP projection bypass. Update ssh-exec instructions and document the complete tool policy inventory.
Keep compact expansion controls left aligned and display the AgentKit truncation marker above them on both platforms. Cover narrow and wide native layouts, both button states, and marker ordering.
An adversarial audit of the three test targets rejected most pruning
candidates: 110 of 183 proposed findings either cited files that do not
exist, mistook architectural guards for tautologies, or treated tests
that drive distinct production branches as duplicates.
Delete tests that pinned no behavior beyond a fixture literal, an
initializer assignment, or an assertion that could not fail, and drop
the unused pluralForms helper. Merge the remaining duplicates into
parameterized tests so each branch keeps its own assertion and failure
localization.
Keep mobile list snapshots and section identity stable during swipe closure.
Move managed-session termination to app-owned background cleanup, transfer existing SSH connections, and allow one silent reconnect when needed. Preserve passive teardown behavior and prevent reopened workspaces from attaching to sessions being terminated.
Cover connection ownership, cleanup cancellation, saved workspace closure, and list updates with regression tests.
Split command execution into run, output, stdin, cancel and list tools for the assistant and MCP. Preserve bounded output tails and use conversation-scoped handles and secret storage with AgentKit 1.20.0.
Add the built-in ssh-exec skill with timestamped log redirection and text/secret stdin workflows. Expand long tool output without adjacent copy controls, update localized presentation, and refresh regression coverage and the website tool name.
BREAKING CHANGE: Replace ssh_exec and ssh_wait with ssh_exec_run and ssh_exec_output. Move stdin to ssh_exec_stdin and remove execution timeout and cwd arguments. Use run_in_background to control foreground execution.
Drop assertions that could not fail in a meaningful way or that only
restated copy already validated by Scripts/check-localization.py.
- Screenshot suites asserted arbitrary PNG byte counts, which AGENTS.md
forbids; the opt-in Attachment.record exports stay.
- AgentSkillTests compared a snapshot revision with itself.
- MetricsCPUUsageTests asserted point.cpu, an alias of cpuUsage.utilization
already pinned by the sibling assertion in that test.
- TerminalTransportTests locked enum declaration order and localized
display names; symbol distinctness now compares against allCases.count
instead of a magic 4.
- AppLanguageTests transcribed catalogs that check-localization.py already
enforces for coverage, format arguments, plurals and spacing. Locale
routing, migration and fallback tests remain, with the Chinese fallback
checks rewritten as relationships rather than literal copy.
Replace the Apple silhouette with the unmodified macOS wordmark from
Simple Icons 16.29.0. Update generator and provenance, retain the stable
HostOS-macos asset/detection IDs, and add a real-bundle source regression.
Other 45 assets and all license files remain byte-identical.
Validation: 19 macOS and 20 iOS focused tests passed; icon integrity,
formatting, AppStrings, localization, and diff checks passed.
Full integration before this isolated resource change: all 1206 shared
iOS tests passed. Two inherited hosted-iOS identity/artwork tests remain
failing. Two unchanged macOS Keychain/Secure Enclave tests return -25308
(errSecInteractionNotAllowed), including a narrow retry. Neither default
full scheme is claimed completely green; no security checks were relaxed.
Remove six unused title/discovery resource APIs and catalog entries while
retaining legacy preference compatibility and all remaining translations.
Replace stale negative discovery assertions with observable row/actions.
Provide the required AppModel to two direct metrics fixtures, sharing their
hosts/model and cleanup. Keep production ownership and UI opt-ins intact.
Validation: 38 focused layout/localization tests, strict formatting,
AppStrings validation, and localization checks passed.
Validate configured MAC addresses in desktop/mobile editors and ordinary,
managed, and custom-identity save paths before credential writes. Preserve
the Auto Wake preference while editing and add localized invalid-address
feedback with regression coverage for rejected writes.
Keep production artwork observation unchanged. Diagnostics confirmed an
inherited hosted-test static-link identity issue; temporary linker and
trace experiments were removed without weakening behavioral assertions.
Validation: 83 focused macOS tests passed; iOS MAC cases passed while the
two known hosted-test identity/artwork failures remain.
Distinguish SSH route admission from Telnet target admission and wire
Telnet access and automatic Wake hooks. Await entitlement resolution for
manual Wake and duplication before host or credential mutation.
Use explicit protocol-extension forwarding overloads so existential
single-argument calls cannot bypass the concrete admission witness.
Remove dead paywall state and update obsolete quota/date test fixtures.
Validation: 81 focused macOS and 58 focused iOS tests passed.
Detach managed sessions when native mobile windows are discarded, retain
live reopen references, and fence ordinary startup commands by input and
session generation. Handle noexec helpers, isolate tmux discovery locale,
and count only verified live contexts against the journal limit.
Move actual shell execution coverage into the hostless framework suite.
Validation: 76 focused macOS tests passed; iOS multiplexer regressions
passed with the inherited hosted-test artwork failure still present.
Make AI, MCP, Metrics, and Wake-on-LAN free while retaining iCloud sync as a Pro feature. Limit free accounts to three saved hosts and preserve excess hosts with remote-operation locks.
Gate host creation, import, and runtime access; preserve stable host admission across persistence and sync. Add upgrade states on macOS and iOS, open the paywall from locked Wake actions, and use wrench icons for MCP.
Include regression coverage, opt-in UI review fixtures, localized copy, and the updated pricing website revision.
Add local Bell and screen-awake controls with per-session bell observation and scene-aware effects. Provide foreground Connection Journey success haptics independent of terminal preferences.
Replace the notification sound with an original bell resource, make AudioToolbox completion queue-safe, and group mobile theme settings.
Extract multiplexer dialect and resumable terminal engine registries while preserving session and checkpoint compatibility.
Share local-network and theme policies, add regression coverage, and update discovery actions to Cancel and Rescan with localized Telnet copy.
Add SSH-authenticated transports with independent settings, protocol-aware menus and snippets, and tmux session persistence.
Restore eligible iOS workspaces from encrypted transport checkpoints without adding a macOS recovery entry. Include schema migrations and regression coverage.
Kaede Akino