Added support for excluding custom messages from model context and compaction or branch summaries, while allowing queued steering and follow-up messages to continue correctly after excluded messages. Added a `prewarm` stream option for Anthropic Messages requests to refresh prompt caches without generating billable output; it is rejected for budget thinking, forced tool choice, and structured output. Claude usage reports now include the account subscription plan in `metadata.subscriptionPlan`, supporting Max 5x, Max 20x, Pro, Team, and Enterprise plans. Added `retryUnreachableBroker` and `onBrokerRetry` options to `discoverAuthStorage`, allowing hosts to control startup retries for temporarily unavailable authentication brokers and display retry status. Added support for prompt-cache prewarming on the first-party Anthropic endpoint via the `isPromptCachePrewarmSupported` compatibility capability. Added built-in MCP transforms, including the Codex Desktop node_repl replacement. Built-ins can be disabled with mcp.builtinTransforms; /mcp list identifies replaced servers and /mcp enable node_repl can explicitly connect one. Added optional MCP eval namespaces through mcp.evalNamespaces, exposing servers as JavaScript and Python namespaces with documentation at xd://eval/. Results preserve text and structured data, and binary resources are saved as local:// files. Added model-generated names for /tan agents and their tray cards, with a readable random fallback when naming fails or times out. Added configurable cache-warming strategies (auto, keepalive, and compact), including long-tool-call keepalive and pre-expiry prompt compaction. Cache-warming events now report the selected method and costs. Added optional cross-instance task discovery and messaging with task.crossInstance, exposing same-folder instances through agent://peers. Added btw mode: drafts beginning with ? submit as /btw questions and receive distinct UI indication. Added Tern live cards for running subagents and background jobs, including progress, tool counts, status, controls, transcripts, and restored access to finished agents. Added daemon update survival and optional TCP hosting through OMP_DAEMON_TCP, preserving running services, PTYs, and language servers across daemon upgrades. Loaded speech-to-text and text-to-speech models once per machine, unloading them after 15 minutes of inactivity. Added ovnfs worktree support to /wt, the project resume picker, and omp worktree list/clear, while preserving session resume from the source repository after checkout removal. Added Claude subscription plans to omp usage and /usage output. Added progress reporting for background skill-description and MCP namespace hint generation. Added excludeFromContext to pi.sendMessage payloads so extensions can display and save per-turn messages without sending them to the model. Added a 30-second retry window for CLI and print-mode commands while the authentication broker restarts or is briefly unavailable. Added a notify option to the bash tool for reporting matching output from background commands and named services, waking idle sessions or ending blocked waits. Added `VcsGitRepo.worktreeToGit()` to convert worktree content into the blob representation produced by `git add`. Added support for using a `Shell` as a process's shell, routing commands through the process's standard streams while preserving the process session and terminal for external commands. Added `ShellRunOptions.speech` to connect the host speech handler to the shell's `say` command. Added an in-process `file` shell builtin with broad file-type detection, MIME and encoding output, archive/image/document support, and compatibility with the commonly used `file` command options. Expanded the Overview's Token mix to include off-transcript model calls—such as title generation, job captions, judge queries, and cache warming—with per-source request counts, token totals, and usage shares. Added a Decision models card to the Overview showing requests, tokens, and usage share by judgment use case for native judgment models. Request details for title-generation and judge calls now display the generated reply or recorded judge answers. Added the `btw` composer mode, including the `btwMode` theme color in all bundled themes. Added feature-gated native docking for runtime children, allowing components to appear only on terminals that advertise the required capability. Added options to preserve intraword underscores when stripping inline Markdown and to open `JobsSheet` on a selected job. Custom message payloads can now be excluded from model context while remaining available in normalized and restored saved entries. Added transcript blocks for output reported by running background jobs and services. Added utilities for formatting and parsing session spill URLs using the `local://N.log` format. Added `postmortem.exitCleanlyOnStopSignals()` so long-running services exit successfully after completing cleanup on SIGINT or SIGTERM. Added tray state and interaction properties, including focus, peek, quiet, main status, item references, card status and actions, and event state. Added layout transition properties to nodes for coordinating layout animations. Added support for folding a dock composer into its compact row.