perf: cut startup cost across natives, providers, plugins and MCP (W1-W6) - #3846

Avoid fs reads for bundled skill content

When this runs from a compiled Bun executable, sourcePath resolves under /$bunfs/...; literal text imports are available there, but readFileSync/Bun.file against that path returns ENOENT. Because content on the exported bundled default definitions and embedded skills now calls this getter, any compiled-binary caller that still reads definition.content/skill.content instead of the new async loadContent() loses the bundled workflow skill body. Keep the text behind the generated import loaders/cache rather than reading the source path synchronously.

AGENTS.md reference: AGENTS.md:L112-L122

Useful? React with 👍 / 👎.

Install the supervisor SIGTERM guard before lazy import

For the internal --internal-managed-owner-supervisor invocation, this lazy import leaves a startup window before managed-owner-supervisor.ts's module-scope SIGTERM capture is installed. If the pane close flow sends SIGTERM while the CLI is still loading or before this branch imports the module, the supervisor exits via the default handler instead of preserving bootstrapSigtermPending and relaying/observing the exact child termination, so managed-owner recovery loses its trusted terminal evidence. Keep a tiny early handler or load the supervisor guard before normal CLI initialization.

AGENTS.md reference: AGENTS.md:L101-L110

Useful? React with 👍 / 👎.

Dispatch each pooled notification only once

When a shared MCP entry receives a notification, this calls the first acquirer's onNotification directly and then emits the same event to every lease, including that first lease. Its manager therefore processes tools/list_changed, resources/list_changed, and similar events twice, causing duplicate refresh requests and duplicate consumer callbacks whenever it owns the physical connection; route notifications exclusively through the lease fan-out.

Useful? React with 👍 / 👎.

Establish artifacts before pruning ephemeral sessions

For a non-persistent/SDK session with no session file and no previously saved artifact, getArtifactManager() returns null; consequently nothing is added to published, the committed plan is empty, and stale tool outputs are never pruned. Long ephemeral sessions therefore fall through to compaction or context overflow even though ensureArtifactManager() can create the supported temporary artifact store; establish that store before publishing prune candidates.

Useful? React with 👍 / 👎.

Reinstall SDK transport handlers after restart

When the same SessionSdkSessionRuntime is stopped and then started again, these one-time disposers permanently remove the connection-close, capability-negotiation, and malformed-frame handlers, while start() never registers replacements. The restarted runtime consequently retains reverse leases after disconnects, ignores negotiated capabilities, and stops returning structured malformed-frame errors; either make the runtime explicitly terminal after stop or recreate these subscriptions on every start.

Useful? React with 👍 / 👎.

Keep nullable plugin parameters required

For plugin tools whose schema is inferred from source, a declaration such as z.string().nullable() is classified as optional here. Nullable fields may contain null, but their presence is still required, so the published schema permits the model to omit a value that the plugin's runtime validator rejects; the static expression parser also reduces this chained expression to the underlying string schema and therefore rejects the valid null value. Treat only optional wrappers as removing a property from required, while preserving nullability in the property schema.

Useful? React with 👍 / 👎.

Preserve best-effort workspace scanning in lazy mode

With workspaceTree.mode="lazy", any native scan error or filesystem failure from the service now escapes this first-turn barrier and rejects prompt() instead of omitting the workspace tree as the fallback branch and the previous implementation do. Because LazyService retains the failed state, subsequent prompts keep rejecting as well, making an optional context scan permanently disable the session; catch scan/service failures here and continue with an undefined tree.

Useful? React with 👍 / 👎.

Prevent transport startup from winning after shutdown

If stop() runs while transport.start() is pending, the transport shutdown path waits for that start and then closes the newly created server, but this continuation still sets #transportStarted to true and returns its endpoint. The start caller can therefore receive a URL whose endpoint file/server has already been removed, while the runtime incorrectly reports the transport as started; coordinate start and stop with a generation or shutdown flag so an in-flight start cannot publish success after shutdown begins.

Useful? React with 👍 / 👎.

Replace inline backend imports with approved lazy loaders

Replace this await import() path, along with the other method-level imports in these backend facades, with the repository's approved top-level/static loading pattern. The repository contract explicitly prohibits inline imports, and these calls make the memory backend's dependency and compiled-binary loading behavior depend on runtime method execution rather than the statically inspectable module graph.

AGENTS.md reference: AGENTS.md:L103-L106

Useful? React with 👍 / 👎.

Account for artifact references when admitting a prune

When below-threshold pruning is near minimumSavings, this plan is built without artifactRefMaxChars, so admission assumes the replacement has no artifact URI; the method then appends the published URI and can commit even after actual savings fall below the configured minimum, because the optional commit gate checks only cache-reset cost. This can trigger a history rewrite and provider-cache reset for a prune that the documented minimum-savings gate should have rejected; build the committed plan with the same artifact-reference budget used by the preflight estimate or recheck the final savings against the minimum.

Useful? React with 👍 / 👎.

Register SDK-only sessions with the broker

When notifications are disabled—the default for ordinary top-level sessions—createAgentSession selects this SDK-only extension, but startup ends after runtime.start() and never calls the available runtime.registerWithBroker(...) path. Unlike the previous notification-host path, no host_registered record reaches SessionIndex, so broker-backed session.list, SDK attach, and relay commands cannot discover these otherwise-live endpoints.

Useful? React with 👍 / 👎.

Include the session identity in SDK endpoint records

For sessions using the new WebSocket transport, this endpoint record omits sessionId. Broker endpoint resolution explicitly rejects records unless endpoint.sessionId === record.sessionId (sdk/broker/broker.ts:784-791), and lifecycle readiness applies the same check, so even after the SDK-only host is indexed, broker attach/readiness treats its endpoint as stale. Publish input.sessionId in the discovery JSON, matching the existing endpoint contract.

Useful? React with 👍 / 👎.

Load syntax bindings before one-shot rendering

On non-macOS hosts and in non-interactive/print mode, theme initialization does not preload the native bindings. The first code block therefore captures undefined, starts an asynchronous load, and immediately returns unhighlighted text; print mode has no subsequent render, so syntax highlighting remains absent despite syntaxHighlighting.enabled defaulting to true. Await the binding load before one-shot rendering or trigger a rerender when it completes.

Useful? React with 👍 / 👎.

Cap explicit artifact ranges before reading

For an artifact larger than 16 MiB, an explicit open-ended or wide query such as ?range=0- enters this branch and materializes the entire file, completely bypassing the ceiling applied to bare reads. Large spilled artifacts can therefore allocate hundreds of MiB or more and crash the session; clamp every requested range to a bounded maximum before calling text().

Useful? React with 👍 / 👎.

Cancel the SDK shutdown timeout after a fast stop

When current.stop(true) resolves before 250 ms, Promise.race drops the timeout promise but does not cancel its timer. Bun keeps that timer alive, so an otherwise-complete SDK shutdown can hold process exit open for the remainder of the 250 ms window; use a cancellable timeout and clear it when the stop settles.

AGENTS.md reference: AGENTS.md:L108-L109

Useful? React with 👍 / 👎.

Discard stale asynchronous history navigation

When history has not loaded yet, pressing Up starts this asynchronous load; if the user types or otherwise changes the draft before it resolves, the continuation still calls #navigateHistory(direction) and replaces the newer draft with an old prompt. Capture the editor/document state at the initial keypress and only navigate if it is still unchanged when loading completes.

Useful? React with 👍 / 👎.

Exclude unavailable Telegram tools from discovery

When tools.discoveryMode uses its default all value and Telegram notifications are disabled, telegram_send falls through to this unconditional true, so createTools() publishes a deferred LazyAgentTool without running TelegramSendTool.createIf(). The model can then discover and activate the tool, but its first execution loads a null implementation and fails with the generic “failed to load” error. Mirror the provider-enabled predicate from tools/telegram-send.ts in descriptor availability so the discovery catalog does not advertise an unavailable capability.

Useful? React with 👍 / 👎.

Decouple theme file watching from appearance detection

When an interactive user sets theme.watchFiles=false, this passes false as initTheme's general enableWatcher argument. That argument also gates startSigwinchListener() and the macOS appearance observer in theme.ts, so disabling custom-theme file reloads unexpectedly disables automatic dark/light theme detection and subsequent terminal appearance updates. Keep appearance monitoring enabled for interactive sessions and gate only the file watcher with this setting.

Useful? React with 👍 / 👎.

Reuse one pool across MCP managers

When multiple MCP managers in the same process connect to a server configured with sharing: "shared", each manager still creates its own MCPConnectionPool, so their keys and leases can never meet and each facade opens a separate physical connection. I checked the production constructors in sdk/session.ts, the MCP loader, and runtime command controllers; they omit options.pool, while only tests inject a common pool. Use a process/session-host-owned shared pool for managers that are expected to honor shared mode.

Useful? React with 👍 / 👎.

Honor history disablement after storage loads

If history is initially enabled and the editor loads HistoryStorage, changing history.enabled to false later has no effect because this cached-storage return bypasses the setting check; subsequent prompts continue to be written even though history search is disabled. The same issue occurs when the setting is turned off while openAsync() is pending, because its continuation still installs the storage. Recheck the live setting before returning/installing storage and detach persistence when the setting is disabled.

Useful? React with 👍 / 👎.

Replace inline imports in the built-in tool loader

This new loader table resolves every built-in through runtime import() calls, despite the repository contract requiring top-level imports only. Besides making dependencies invisible to the static module graph, this leaves compiled-binary loading behavior dependent on runtime activation; use the repository's approved statically imported lazy-loader pattern instead.

AGENTS.md reference: AGENTS.md:L103-L106

Useful? React with 👍 / 👎.

Preserve frontmatter-free skills during lazy scanning

When requireDescription is false, a SKILL.md without YAML frontmatter was previously loaded using its directory name and full body, but readSkillFrontmatter() returns null for that valid input and this branch now drops it unconditionally. The Codex, OpenCode, Claude, and agents discovery paths call this scanner without requiring a description, so existing frontmatter-free custom skills silently disappear from discovery.

Useful? React with 👍 / 👎.

Hide SSH until a host configuration exists

With the default tools.discoveryMode="all", this discoverable loader publishes ssh even when the session has no configured SSH hosts. Activating it then makes loadSshTool() return null, which LazyAgentTool surfaces as the generic Tool "ssh" failed to load error; the previous eager factory path omitted the tool in this case. Check host availability before adding the deferred descriptor, or keep SSH eager until hosts are known.

Useful? React with 👍 / 👎.

Hide recipes when no runnable tasks exist

Because recipe discovery and recipe.enabled both default on, repositories without a supported runner containing tasks still advertise this deferred tool. On activation, RecipeTool.createIf() detects zero runnable tasks and returns null, so the model receives a generic failed-to-load error for a capability that should not have been discoverable; preserve the factory's task-detection guard before publishing the descriptor.

Useful? React with 👍 / 👎.

Declare the lazy initialization result type explicitly

Replace this ReturnType<>-derived alias with an explicit initialization-result interface. The repository contract prohibits ReturnType<>, and keeping the result shape explicit prevents this service's internal state type from changing implicitly whenever the initializer callback signature is edited.

AGENTS.md reference: AGENTS.md:L104-L104

Useful? React with 👍 / 👎.

Convert the pool helpers to ES private methods

Convert this and the other newly introduced private pool helpers to #waitForPendingEntry, #closeEntry, and so on. The repository requires ES #private fields rather than TypeScript access modifiers, so the current pool implementation violates the established runtime-private convention throughout the class.

AGENTS.md reference: AGENTS.md:L108-L108

Useful? React with 👍 / 👎.

Preserve strictness in inferred plugin schemas

When a source-inferred plugin declares parameters: z.object({...}).strict(), callName() parses only the initial object() call and this emits additionalProperties: true, ignoring the trailing strict wrapper. The registry consequently advertises extra keys as valid even though the loaded Zod schema rejects them at execution time, so model-generated calls that conform to the published schema can fail validation; preserve supported trailing constraints or reject unsupported chains instead.

Useful? React with 👍 / 👎.

Publish SDK endpoints without following existing symlinks

When a resumed session's known endpoint filename has been pre-created as a symlink, this direct writeFile follows it and truncates a file outside .gjc/state; the subsequent chmod also operates on that target. Because the endpoint path is workspace-local, stale or hostile workspace state can therefore corrupt arbitrary user files during SDK startup; publish a mode-0600 temporary file and atomically rename it over the endpoint, as the other SDK discovery writers do.

Useful? React with 👍 / 👎.

Cancel pending macOS assertion acquisition on release

When a turn exits synchronously after #beginInFlight()—for example, the preflight cancellation or missing-model checks can throw before the first await—#releasePowerAssertion() runs while #powerAssertion is still unset, and this queued continuation subsequently starts the assertion anyway. The orphaned assertion can prevent system sleep until another turn finishes or the session is disposed; keep acquisition synchronous or track cancellation/release state before calling MacOSPowerAssertion.start().

AGENTS.md reference: AGENTS.md:L105-L105

Useful? React with 👍 / 👎.

Preserve symlinked plugins during legacy discovery

When an existing user or project plugin is installed as a symlink and has not yet been indexed in registry.json, this filter skips it because Dirent.isDirectory() is false for symlinks. The previous discovery path explicitly accepted isSymbolicLink(), so these valid legacy plugins now silently disappear instead of being migrated and activated; retain symlink candidates and let the compiler's containment and hash validation decide whether they are safe.

Useful? React with 👍 / 👎.

Reinitialize memory service when the backend setting changes

After session startup initializes this lazy service, changing memory.backend through the interactive settings selector cannot change its cached value. For example, a session started with local continues injecting and running the local backend after the user selects off, whereas the previous per-operation resolver observed the current setting. Reset or replace the service when this setting changes, or make the facade dispatch against the live backend selection.

Useful? React with 👍 / 👎.

Apply syntax-highlighting changes to the live theme

When an interactive user changes syntaxHighlighting.enabled in the settings selector, the new value is persisted but rendering does not change because the theme module copies it into syntaxHighlightingEnabledState only during this startup call, and the selector's change handler has no case that updates that state. Consequently the new appearance toggle has no effect until the CLI is restarted; wire the setting change to a live setter or theme reinitialization.

Useful? React with 👍 / 👎.