@sema-agent/core 5.6.0 → 5.8.0

package/CHANGELOG.md

@@ -1,5 +1,60 @@
 # Changelog
 
+## 5.8.0 (2026-08-04)
+
+_The limits restructure: time leaves the task-limit axis, tokens become the primary budget, and every default is empty. One release, five construction stages — there is no intermediate version. Old usage fails loudly (compile errors for removed fields; a typed refusal for unknown limit keys at runtime) — nothing is silently ignored._
+
+### BREAKING — task limits (`TaskSpec.limits` is the one budget surface)
+
+- `limits` now holds every budget knob: `maxTokens` (primary, cumulative per-slice window — promoted from report-only to a GATE) / `maxCostUsd` / `maxTurns` / `maxWalltimeMs` (opt-in, per-slice active clock, **milliseconds**) / `maxOutputTokens` (unchanged, per-request shape) / `approachNotice` / `budgetStreamCancel` / `degrade` / `brainCallGuardrailMs`. The former top-level `maxTokens` / `maxCostUsd` / `budgetStreamCancel` / `degrade` fields are gone.
+- `limits.timeoutSec` is REMOVED (unit and semantics changed → `maxWalltimeMs`). An unknown key on `limits` is refused at prepare with `config.limit_unknown_key` naming the replacement and the unit; a non-finite or negative value refuses with `config.limit_invalid`. Never silently ignored.
+- EVERY default is empty: `DEFAULT_MAX_TURNS=1000`, `FORK_DEFAULT_MAX_TURNS=200`, the session-scoped background child's injected 30-minute wall clock, the resource-suspend `maxSlices=100` default, and the workflow per-agent 300s cap are all retired. No limit is armed unless configured. Anti-zombie coverage for session-scoped background children moved to the stale-running reaper (`TaskRegistry.reapStaleSessionBackgroundAgents` — collects by STALL, not by wall age). Honest consequence, stated plainly: a zero-config task with a non-terminating brain loop now has no task-level backstop — the engineering guardrails (brain-call guardrail, stall detection) keep it from wedging silently, and the deployment governance windows are the burn ceiling.
+- To-limit behavior is two-tier: WITHOUT `resourceSuspend`, hitting any limit is a loud terminal failure — `TaskResult.status` loses `"timeout"`; all four axes end as `failed` with `limits.max_tokens_exceeded` / `limits.max_cost_exceeded` / `limits.max_turns_exceeded` / `limits.max_walltime_exceeded` (the code names the AXIS, replacing `limit.timeout` / `limit.max_turns` / `budget.exceeded` / `budget.precall`), salvage attached. WITH `resourceSuspend` (+ checkpoint infrastructure), a per-slice window exhaustion suspends with a checkpoint and resume grants the next window; only the optional frozen totals (`resourceSuspend.totalTokens` / `totalBudgetUsd`) are hard stops. `ResourceLimitReason` splits `"budget"` into `"budget_tokens"` / `"budget_cost"` — consumers pinning either closed set are affected and named here.
+- Priority when several axes trip at one boundary: tokens > cost > turns > walltime (attribution only — every configured axis stays armed).
+- `limits.approachNotice` replaces `deadlineNudge`: one "start converging" frame at ~80% and one "deliver now" at ~95% of the tightest configured budget, measured in BUDGET (tokens/cost/turns), never in clock; one-shot; `stats.mechanisms.approachNoticesSent`; a continuation gate keeps the notice from extending a run the model had already ended.
+
+### BREAKING — retired mechanisms
+
+- The design/128/130 deadline family is gone: `deadlineNudge` / `callCapByDeadline` / `gracefulFinalize`, the write-out cushion mathematics, and the per-call deadline shrink.
+- tool-cut is gone end to end: `stats.mechanisms.{toolCuts,toolCutKills,pendingCutTools,finalizeInjected,nudgesSent,capShrinks,callCutoffs,toolClamps}`, `engineCut:"walltime_writeout"`, the bash exec deadline clamp, and `cut-kill.ts`.
+- The brain soft-cutoff is gone: `errorKind:"walltime_cutoff"`, `WALLTIME_CUTOFF_MESSAGE`, `StreamOptions.callDeadlineMs`, the loop's `walltime_cutoff_recovery`.
+- auto-compaction's walltime gate / abort race / throughput EWMA are gone (`clampReason` loses `"walltime"`).
+- Stall detection (connect / first-token / idle) is DECOUPLED from the retired walltime lane and now arms unconditionally (guardrail class; constants renamed `STALL_*`; knob semantics unchanged).
+- The `ExecutionEnv.exec` `onSpawn` / `onDetachAdopted` seam is retired (its only consumer left with tool-cut; zero producers remained).
+- The Terminal-Bench surface is retired: the `sema-tb` bin, `tb-env`, the `tb/` adapter tree, and all thirteen `TB_*` env knobs. No switch survives.
+
+### BREAKING — checkpoint / ledger / config
+
+- `ResourceLedger.totalWalltimeSec` and `remainingWalltimeMs()` retired; `totalTokens` + `remainingTokens()` added; `spentWalltimeMs` REMAINS (accounting only). Checkpoint version 5 (`TOKEN_CHECKPOINT_VERSION`); a v4 checkpoint carrying the retired wall-clock allocation is refused pre-CAS with `checkpoint.walltime_axis_retired` (the checkpoint stays pending — finish it on the previous release, or start fresh; no silent migration). `resourceSuspend.totalWalltimeSec` → `resourceSuspend.totalTokens`.
+- Config catalog: `limits.{timeoutSec,deadlineNudge,callCapByDeadline,gracefulFinalize}`, `budget.{maxCostUsd,maxTokens,totalWalltimeSec}`, `delegation.{forkMaxTurns,sessionBgTimeoutSec}` retired/moved; `limits.{maxWalltimeMs,approachNotice,maxTokens,maxCostUsd}`, `budget.totalTokens` added. Workflow script `limits` whitelist: `{maxTurns,timeoutSec}` → `{maxTurns,maxWalltimeMs,maxTokens,maxCostUsd}`; `WorkflowChildCaps.perAgentTimeoutSec` → `perAgentMaxWalltimeMs` (no default).
+- Workflow resume journals from 5.7 cache-miss entirely on this version (the spec shape participates in the call identity) — reruns, not errors.
+- Trace: the eight mechanisms fields above are gone; `timerLatenessMs` / `walltimeSyncBackstopFired` REMAIN (the per-slice wall timer survives as `maxWalltimeMs`, so its starvation evidence keeps a producer).
+
+### Fixed
+
+- memory-engine: partition adoption is a single atomic rename (the delete-then-rename window could destroy a concurrently adopted partition and then disclose the opposite of what happened); fileless-scaffolding cleanup uses `rmdir` only, so a peer's files are unremovable by construction. Both control-plane lock waits are bounded and yield every round — a permanent errno can no longer starve the event loop in a silent spin (the fail-open sidecar lock now honors its documented "proceed anyway" by going lockless after three post-deadline steal rounds).
+- cc stores: the mailbox sweep AND `drop` now judge and delete inside the box's cross-process lock (a compliant writer's freshly committed message can no longer be deleted unseen, and another holder's lock directory is never torn down); the cc lockfile's retry bound is read before every round (three `continue` shortcuts skipped it — a permanently failing reclaim spun forever); the lease table's entry keys fold the same way the table itself does (two store instances over one directory via a symlink could each grant the same lease).
+- session policy: the tighten gate's path-containment fold is win-family-conditional, matching the runtime's — a POSIX sibling directory whose literal name embeds backslashes no longer reads as "within" the confined root (a loosening the gate exists to refuse). The transcript-dir gate's write-tool literal gains a mechanical sync pin against `PATH_WRITE_TOOLS`.
+- `tool-name-aliases` docstring named a compile-time enforcement symbol that never existed; it now states the real division of labor (the prepare-time audit escalates, the compile pass deliberately does not).
+
+## 5.7.0 (2026-08-04)
+
+_Three collected cars: a parked question is answered by the decision itself (behavior face — see the consumer note), a descendant's terminal notification anchors on the delegation tree, and the env scrub stops deleting infrastructure key names (behavior face)._
+
+### Fixed
+
+- A descendant background agent's terminal notification now anchors on the delegation tree instead of dying with its spawner. A background agent notifies through the injector of the run that spawned it, and that injector parked on that run's OWN session once its lane was gone — so a grandchild launched from a delegated run (root → child A → background grandchild B; A completes, B settles later) parked its terminal on a throwaway child session that never runs again, and the root conversation was never told (the data was never lost — `TaskOutput` still answered; the gap was purely the push face). A delegated agent's TERMINAL frame now also escalates one hop up the delegation chain: a live ancestor takes it in its current run, a dead one repeats the decision one level higher, and the walk ends at the tree root; without a chain link the recorded root anchor is used directly. Strictly additive: the own-session park is kept (a retained child session later resumed still receives what it always received), the live lane is untouched, and every other notification class (shell completions, SendMessage deliveries, external `notify()` events) stays addressed to its own session. A delegated run that can reach neither an uplink nor a root anchor still parks (fail-open) and discloses once (`degraded` / `descendant-terminal-unrouted`).
+
+- The env scrub no longer deletes infrastructure key names. `SECRET_ENV_RE`'s bare `KEY` word matched `PARTITION_KEY` / `SORT_KEY` / `RANGE_KEY` / `IDEMPOTENCY_KEY` / `CACHE_KEY` / `ROUTING_KEY`, and `scrubSecretEnv` DELETES what it matches — a spawned child silently lost variables that carry no credential (writes to a wrong partition, a routing key nothing binds). The word itself cannot be withdrawn (`API_KEY` / `SECRET_KEY` / `PRIVATE_KEY` / `ACCESS_KEY` are the bulk of real credential names), so the fix is the mirror of the existing positive exact-name list: `NON_SECRET_ENV_EXACT_NAMES`, exact names (never a shape — a pattern exemption would be a bypass) confirmed non-credential, consulted before the shape rule with the credential list read first (a name on both lists resolves to the credential reading). An exempted entry is kept and mints no finding. Decorated variants (`USERS_SORT_KEY`, `SORT_KEY_2`) are still dropped. **Consumers pinning the old shape**: a probe asserting `isSecretEnvKey("SORT_KEY") === true`, or that a scrub drops one of the six exempt names, now reads the opposite answer.
+
+### Added
+
+- A durable question can be answered by the decision itself. `ResumeOutcome`'s `policy_ask` arm takes an optional `answer` (a `QuestionAnswer`) when the checkpoint's pending call is the reserved `AskUserQuestion` tool; the resume binds it as the resumed leg's answering face, so the pending call executes against the operator's actual selections. This is the same wiring the documented live arm builds by hand (`resume(token, outcome, { ...config, onQuestion: async () => answer })`) — but minted from the payload, which is the only route available to a caller redeeming an **offline background child**: the revive drive rebuilds that leg's config from the parked row and has no seat to put a closure in, so approving a parked question used to feed the model a fabricated "no human is available" default. The answer is untrusted wire data throughout: it goes through the existing answer fence unchanged (`selected ⊆ options` for the trusted line, off-list values and free-text `note` in a data fence), so multi-select and "Other" free text behave identically on both arms. The binding is one-shot — it answers the decided call only; a new question on the resumed leg is not fed the old answer. The answer is also recorded on the persisted winner, so an `env_failed` reopen must replay the identical answer (a replay may not keep the verdict and swap the text).
+
+### Changed
+
+- **Behavior face** — approving a durable question without an answer is now refused instead of consumed. A `policy_ask` `allow` whose pending call is the reserved question tool is rejected **pre-CAS** (`checkpoint.invalid_outcome`, `detail.field: "answer"`, the checkpoint stays `pending` and redeemable) when it carries neither an `answer` nor a live answering face — where "live" excludes the reserved `QUESTION_AWAITS_RESUME` placeholder, whose contract is that it must never run. Previously that shape consumed the human's approval and let the run finish on a fabricated default (or, with the placeholder, on its config error) with nothing left to redeem. A resume that supplies its own real `onQuestion` is unaffected; so is every non-question approval. Symmetrically, an `answer` attached to a decision that cannot carry one (a `deny`, or a pending call that is not the question tool) is rejected the same way rather than silently dropped. **Consumers pinning the old shape**: a probe that approved a parked question with a bare allow now gets a typed rejection — supply the answer on the outcome, or deny.
+
 ## 5.6.0 (2026-08-03)
 
 _A hardening tail: numeric knobs refuse what they cannot honor, the binary gate reads bytes instead of trusting statistics, and the MCP SDK takes its last zero-cost step before the v2 jump._

package/dist/agents/cascade.d.ts

@@ -6,7 +6,7 @@ export type GateVerdict = boolean | {
 };
 export interface CascadeRung {
     model: ModelRef;
-    overrides?: Partial<Pick<TaskSpec, "limits" | "degrade" | "systemPrompt">>;
+    overrides?: Partial<Pick<TaskSpec, "limits" | "systemPrompt">>;
 }
 export interface CascadeConfig {
     onNotifyError?: (failure: import("../core/safe-notify.js").SafeNotifyFailure) => void;

package/dist/agents/subagent.d.ts

@@ -1,10 +1,10 @@
 import type { Runner } from "../core/runner/runtask.js";
-import type { AgentDefinition, Model, ModelRef, ToolSpec } from "../core/types.js";
+import type { AgentDefinition, Model, ModelRef, TaskLimits, ToolSpec } from "../core/types.js";
 import { type ExecutionEnv } from "../internal/harness.js";
 import type { RunInternals } from "../core/runner/prepare-task.js";
 import type { TaskNotificationPayload } from "../core/task-notification.js";
-import { FORK_DEFAULT_MAX_TURNS, SESSION_BG_DEFAULT_TIMEOUT_SEC, RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX } from "../config/defaults.js";
-export { FORK_DEFAULT_MAX_TURNS, SESSION_BG_DEFAULT_TIMEOUT_SEC, RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX };
+import { RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX } from "../config/defaults.js";
+export { RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX };
 import { SubagentRetainLedger } from "./retain-ledger.js";
 export type { SubagentStep, SubagentEditedFile } from "./subagent-steps.js";
 export declare function notifyResultField(result: string | undefined): string | undefined;
@@ -101,10 +101,7 @@ export interface SubagentToolOptions {
     systemPrompt?: string;
     name?: string;
     purpose?: string;
-    limits?: {
-        maxTurns?: number;
-        timeoutSec?: number;
-    };
+    limits?: TaskLimits;
     maxDepth?: number;
     retainFailedSessions?: boolean;
     onObserverError?: (err: unknown, info: {

package/dist/agents/subagent.js

@@ -17,7 +17,7 @@ import { addWorktree } from "../core/git-worktree-env.js";
 import { shellQuote } from "../tools/fs/search.js";
 import { BG_AGENT_REAP_STOP_ERROR } from "../core/task-registry.js";
 import { extractErrorCode } from "../brain/errors.js";
-import { FORK_DEFAULT_MAX_TURNS, SESSION_BG_DEFAULT_TIMEOUT_SEC, RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX, RUNNING_AGENT_OBSERVE_EVERY_BEATS } from "../config/defaults.js";
+import { RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX, RUNNING_AGENT_OBSERVE_EVERY_BEATS } from "../config/defaults.js";
 function rollupDelegatedCost(stats, nested) {
     if (stats.costMicroUsd === undefined)
         return undefined;
@@ -29,7 +29,7 @@ function delegatedCostField(stats) {
     const total = rollupDelegatedCost(stats, stats.nested);
     return total !== undefined ? { costMicroUsd: total } : {};
 }
-export { FORK_DEFAULT_MAX_TURNS, SESSION_BG_DEFAULT_TIMEOUT_SEC, RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX };
+export { RETAIN_DEFAULT_TTL_MS, RETAIN_DEFAULT_MAX };
 import { SUBAGENT_RESUME_CAP, SubagentRetainLedger, getOrCreateSessionRetainLedger, ensureSessionReapHook, createResumePrompt, } from "./retain-ledger.js";
 import { recordRosterSpawn } from "./roster-store.js";
 import { ObserverDigestTap, ObserverPairing, createObserverReportToolSpec, markObserverTaskId, unmarkObserverTaskId, isObserverTaskId, ObserverResumeStateError, ObserverStoppedByUserError, observerFramingPrompt, observerSlug, resolveObserverDeclaration, } from "./observer.js";
@@ -224,7 +224,7 @@ export function classifySubagentError(child) {
         ? "rate_limit"
         : code === "server" && (msg.includes("overloaded") || msg.includes("529"))
             ? "overloaded"
-            : code === "limit.timeout" || ((code === "network" || code === "server") && msg.includes("timeout"))
+            : code === "limits.max_walltime_exceeded" || ((code === "network" || code === "server") && msg.includes("timeout"))
                 ? "timeout"
                 : code === "network" || code === "server"
                     ? "network"
@@ -237,7 +237,7 @@ function errorKindClause(c) {
 const FAILED_SESSION_RETAIN_TTL_MS = 15 * 60 * 1000;
 const PARTIAL_FINDINGS_MAX_CHARS = 1200;
 const BG_NOTIFY_DRAIN_WINDOW_MS = 2_000;
-function createBgActivityBeat(parentToolCallId, emitTick) {
+function createBgActivityBeat(parentToolCallId, emitTick, noteActivity) {
     let beats = 0;
     let starts = 0;
     return (e) => {
@@ -248,6 +248,7 @@ function createBgActivityBeat(parentToolCallId, emitTick) {
         if (e.type === "tool_start")
             starts += 1;
         beats += 1;
+        noteActivity?.();
         if (beats === 1 || beats % RUNNING_AGENT_OBSERVE_EVERY_BEATS === 0)
             emitTick(starts);
     };
@@ -395,7 +396,7 @@ export function createSubagentResume(deps) {
                     ...(currentTool !== undefined ? { currentTool } : {}),
                     usage: { toolUses: toolStarts },
                 });
-            });
+            }, deps.registry !== undefined && deps.taskId !== undefined ? () => deps.registry.noteBackgroundAgentActivity(deps.taskId) : undefined);
             stream = childRunner.runTaskStream(resumeSpec, undefined, {
                 ...entry.internalsSnapshot,
                 ...(true
@@ -1149,11 +1150,7 @@ function createSubagentToolNode(opts, depth, excluded, extraToolsBudget) {
             const childSystemPrompt = def?.systemPrompt ?? opts.systemPrompt;
             const childDefaultPersona = childSystemPrompt === undefined && !wantsFork ? SUBAGENT_PROMPT : undefined;
             const defMaxTurns = typeof def?.maxTurns === "number" && Number.isFinite(def.maxTurns) && def.maxTurns > 0 ? def.maxTurns : undefined;
-            const childLimits = defMaxTurns !== undefined
-                ? { ...opts.limits, maxTurns: defMaxTurns }
-                : wantsFork && opts.limits?.maxTurns === undefined
-                    ? { ...opts.limits, maxTurns: FORK_DEFAULT_MAX_TURNS }
-                    : opts.limits;
+            const childLimits = defMaxTurns !== undefined ? { ...opts.limits, maxTurns: defMaxTurns } : opts.limits;
             const nextExcluded = def ? new Set([...excluded, def.name]) : excluded;
             const defAllowsDelegation = def === undefined || toolNameAllowed(opts.name ?? DEFAULT_SUBAGENT_TOOL_NAME, def.allowTools, def.denyTools);
             if (depth + 1 < maxDepth && defAllowsDelegation) {
@@ -1233,7 +1230,7 @@ function createSubagentToolNode(opts, depth, excluded, extraToolsBudget) {
                     enableBlockedReport: false,
                     limits: {
                         maxTurns: typeof observerDef.maxTurns === "number" && Number.isFinite(observerDef.maxTurns) && observerDef.maxTurns > 0 ? observerDef.maxTurns : 8,
-                        timeoutSec: 120,
+                        maxWalltimeMs: 120_000,
                     },
                 });
                 const runObserverLeg = async (spec) => {
@@ -1722,9 +1719,6 @@ function createSubagentToolNode(opts, depth, excluded, extraToolsBudget) {
                     });
                     const forkBgStartedAt = Date.now();
                     const bgForkSpec = { ...buildChildSpec(abort.signal), sessionId: forkedId, requireExistingSession: true, objective: forkObjective };
-                    if (sessionScopedBg && bgForkSpec.limits?.timeoutSec === undefined) {
-                        bgForkSpec.limits = { ...(bgForkSpec.limits ?? {}), timeoutSec: SESSION_BG_DEFAULT_TIMEOUT_SEC };
-                    }
                     const s2ForkNotifyReady = (inject) => {
                         bg.registry.attachAgentNotify(taskId, inject);
                     };
@@ -1744,7 +1738,7 @@ function createSubagentToolNode(opts, depth, excluded, extraToolsBudget) {
                             ...(currentTool !== undefined ? { currentTool } : {}),
                             usage: { toolUses: toolStarts },
                         });
-                    });
+                    }, () => bg.registry.noteBackgroundAgentActivity(taskId));
                     const bgForkInternals = bgSink
                         ? {
                             ...forkInternals,
@@ -2195,9 +2189,6 @@ function createSubagentToolNode(opts, depth, excluded, extraToolsBudget) {
                     ...(forwardDurableApproval ? { durableApproval: { ...ctx.durableApprovalForChildren } } : {}),
                     ...(reviveRow !== undefined ? { requireExistingSession: true } : {}),
                 };
-                if ((sessionScopedBg || reviveRow !== undefined) && bgSpec.limits?.timeoutSec === undefined) {
-                    bgSpec.limits = { ...(bgSpec.limits ?? {}), timeoutSec: SESSION_BG_DEFAULT_TIMEOUT_SEC };
-                }
                 let reviveAttachedResolve;
                 const reviveAttached = reviveRow !== undefined ? new Promise((r) => (reviveAttachedResolve = r)) : undefined;
                 const s2NotifyReady = (inject) => {
@@ -2222,7 +2213,7 @@ function createSubagentToolNode(opts, depth, excluded, extraToolsBudget) {
                         ...(currentTool !== undefined ? { currentTool } : {}),
                         usage: { toolUses: toolStarts },
                     });
-                });
+                }, () => bg.registry.noteBackgroundAgentActivity(taskId));
                 const bgInternals = bgSink
                     ? {
                         ...childInternals,

package/dist/agents/teacher.js

@@ -192,7 +192,7 @@ async function runTeacherCore(runner, studentSpec, teacher) {
             ...helperBase(),
             objective: helperObjective,
             enableBlockedReport: false,
-            limits: { maxTurns: 1, timeoutSec: studentSpec.limits?.timeoutSec },
+            limits: { maxTurns: 1, ...(studentSpec.limits?.maxWalltimeMs !== undefined ? { maxWalltimeMs: studentSpec.limits.maxWalltimeMs } : {}) },
             signal: extSignal,
         });
         await runner.sessions.release(r.sessionId).catch(() => undefined);
@@ -341,7 +341,7 @@ async function runTeacherCore(runner, studentSpec, teacher) {
             ...teacherModelFields(),
             systemPrompt: teacher.prompts?.teacher ?? TEACHER_PROMPT,
             enableBlockedReport: false,
-            limits: { maxTurns: 4, timeoutSec: studentSpec.limits?.timeoutSec },
+            limits: { maxTurns: 4, ...(studentSpec.limits?.maxWalltimeMs !== undefined ? { maxWalltimeMs: studentSpec.limits.maxWalltimeMs } : {}) },
             signal: extSignal,
         });
         await runner.sessions.release(r.sessionId).catch(() => undefined);
@@ -363,7 +363,7 @@ async function runTeacherCore(runner, studentSpec, teacher) {
         else if (result.status === "blocked") {
             trigger = "blocked";
         }
-        else if (result.status === "failed" || result.status === "timeout") {
+        else if (result.status === "failed") {
             trigger = "failed";
         }
         else if (!verifyOutput) {

package/dist/agents/team.d.ts

@@ -56,7 +56,7 @@ export interface TeamDiscussionOptions {
     };
     onEvent?: (e: TeamEvent) => void;
     limits?: {
-        timeoutSec?: number;
+        maxWalltimeMs?: number;
         maxTurns?: number;
         maxTokens?: number;
         maxCostUsd?: number;

package/dist/agents/team.js

@@ -63,7 +63,7 @@ export async function runTeamDiscussion(opts) {
             : undefined;
     const memberLimits = opts.limits
         ? {
-            ...(opts.limits.timeoutSec !== undefined ? { timeoutSec: opts.limits.timeoutSec } : {}),
+            ...(opts.limits.maxWalltimeMs !== undefined ? { maxWalltimeMs: opts.limits.maxWalltimeMs } : {}),
             ...(opts.limits.maxTurns !== undefined ? { maxTurns: opts.limits.maxTurns } : {}),
         }
         : undefined;

package/dist/agents/verify.js

@@ -101,7 +101,7 @@ export async function verifyCompleted(runner, result, specBase, objective, confi
             handsReadOnly: config.verifierHandsReadOnly ?? true,
             outputSchema: VerdictSchema,
             enableBlockedReport: false,
-            limits: { timeoutSec: specBase.limits?.timeoutSec },
+            limits: { ...(specBase.limits?.maxWalltimeMs !== undefined ? { maxWalltimeMs: specBase.limits.maxWalltimeMs } : {}) },
             getApiKeyAndHeaders: specBase.getApiKeyAndHeaders,
             signal: specBase.signal,
         });

package/dist/brain/anthropic.js

@@ -1,9 +1,8 @@
 import {} from "../internal/llm.js";
 import { BrainError } from "./errors.js";
 import { DEGENERATE_MESSAGE, trimDegenerateTail } from "./repetition.js";
-import { createRepetitionPoll, createWalltimeGate, parseStreamedToolArgs } from "./stream-shared.js";
+import { createRepetitionPoll, parseStreamedToolArgs } from "./stream-shared.js";
 import { mintFallbackToolCallId } from "./tool-call-id.js";
-import { WALLTIME_CUTOFF_MESSAGE } from "./walltime.js";
 import { emitBrainTelemetry } from "./status-sink.js";
 import { errorResultMediaNote, IMAGE_OMITTED_NO_VISION, modelSupportsVision, sendableImages } from "./media-degrade.js";
 import { ANTHROPIC_RESERVED, applyExtraBody, stripAuthHeaders } from "./request-params.js";
@@ -233,7 +232,6 @@ export function createAnthropicBrain(config = {}) {
             signal: options?.signal,
             config,
             httpLabel: "anthropic",
-            callDeadlineMs: options?.callDeadlineMs,
             stallTimeouts: options?.stallTimeouts,
             buildRequest: () => {
                 const apiKey = options?.apiKey ?? config.apiKey;
@@ -342,7 +340,6 @@ export function createAnthropicBrain(config = {}) {
                 let degenerate = false;
                 let degenTextBlock;
                 const rep = createRepetitionPoll();
-                const walltime = createWalltimeGate({ callDeadlineMs: options?.callDeadlineMs, isDegenerate: () => degenerate, cancel: () => ctrl.cancel() });
                 let lastDegenCheck = 0;
                 let lastDegenCheckThinking = 0;
                 const detectRep = config.detectRepetition !== false;
@@ -497,10 +494,7 @@ export function createAnthropicBrain(config = {}) {
                             }
                         }
                     },
-                    onDeadline: () => walltime.onDeadline(),
                     onLine(rawLine) {
-                        if (walltime.shouldDropLine())
-                            return;
                         const line = rawLine.trim();
                         if (!line.startsWith("data:"))
                             return;
@@ -576,40 +570,35 @@ export function createAnthropicBrain(config = {}) {
                         const staticReasoningCut = truncatedEmpty && !dynamicCut && reasoningSeen && options?.staticReasoningCutDowngrade === true;
                         const safetyCut = stopReason === "refusal";
                         const contextWindowExceeded = stopReason === "model_context_window_exceeded";
-                        const emptyNoFinish = noUsableContent && stopReason == null && !degenerate && !walltime.cut;
+                        const emptyNoFinish = noUsableContent && stopReason == null && !degenerate;
                         const maxTokensNote = `effective max_tokens=${sentMaxTokens ?? "?"} (from ${sentMaxTokensLane === "options" ? "options.maxTokens" : sentMaxTokensLane === "model" ? "model.maxTokens" : "config default"})`;
-                        const errorMessage = walltime.cut
-                            ? WALLTIME_CUTOFF_MESSAGE
-                            : degenerate
-                                ? DEGENERATE_MESSAGE
-                                : safetyCut
-                                    ? `[refusal] response refused by the model (stop_reason="refusal"${stopDetails?.category ? `, category="${stopDetails.category}"` : ""}) — ${stopDetails?.explanation ?? "the output was withheld for policy and is unreliable"}`
-                                    : contextWindowExceeded
-                                        ? `[invalid_request] model context window exceeded (stop_reason="model_context_window_exceeded") — the conversation no longer fits the model's context window`
-                                        : (toolError ??
-                                            (truncatedEmpty && !dynamicCut && !staticReasoningCut
-                                                ? thinkingRequested
-                                                    ? `[length_empty] response truncated at max_tokens with no answer text (${maxTokensNote}) — extended thinking likely consumed the entire budget; raise max_tokens or lower thinkingBudgetShare`
-                                                    : reasoningSeen
-                                                        ? `[length_empty] response truncated at max_tokens with no answer text (${maxTokensNote}) — the model emitted reasoning although no thinking was requested (a budget-ignoring gateway?) and it consumed the entire output budget; raise max_tokens`
-                                                        : `[length_empty] response truncated at max_tokens with no answer text (${maxTokensNote}); raise max_tokens`
-                                                : emptyNoFinish
-                                                    ? `[stream_torn] model stream ended with no content and no stop_reason — the response was lost (a torn stream or an in-band provider error frame)${malformedFrames > 0 ? ` — ${malformedFrames} unparseable data frame(s) were dropped on this stream (first sample: ${JSON.stringify(malformedSample)})` : ""}`
-                                                    : undefined));
-                        const errored = walltime.cut ||
-                            degenerate ||
+                        const errorMessage = degenerate
+                            ? DEGENERATE_MESSAGE
+                            : safetyCut
+                                ? `[refusal] response refused by the model (stop_reason="refusal"${stopDetails?.category ? `, category="${stopDetails.category}"` : ""}) — ${stopDetails?.explanation ?? "the output was withheld for policy and is unreliable"}`
+                                : contextWindowExceeded
+                                    ? `[invalid_request] model context window exceeded (stop_reason="model_context_window_exceeded") — the conversation no longer fits the model's context window`
+                                    : (toolError ??
+                                        (truncatedEmpty && !dynamicCut && !staticReasoningCut
+                                            ? thinkingRequested
+                                                ? `[length_empty] response truncated at max_tokens with no answer text (${maxTokensNote}) — extended thinking likely consumed the entire budget; raise max_tokens or lower thinkingBudgetShare`
+                                                : reasoningSeen
+                                                    ? `[length_empty] response truncated at max_tokens with no answer text (${maxTokensNote}) — the model emitted reasoning although no thinking was requested (a budget-ignoring gateway?) and it consumed the entire output budget; raise max_tokens`
+                                                    : `[length_empty] response truncated at max_tokens with no answer text (${maxTokensNote}); raise max_tokens`
+                                            : emptyNoFinish
+                                                ? `[stream_torn] model stream ended with no content and no stop_reason — the response was lost (a torn stream or an in-band provider error frame)${malformedFrames > 0 ? ` — ${malformedFrames} unparseable data frame(s) were dropped on this stream (first sample: ${JSON.stringify(malformedSample)})` : ""}`
+                                                : undefined));
+                        const errored = degenerate ||
                             safetyCut ||
                             contextWindowExceeded ||
                             emptyNoFinish ||
                             (toolError !== undefined && noUsableContent) ||
                             (truncatedEmpty && !dynamicCut && !staticReasoningCut);
-                        const errorKind = walltime.cut
-                            ? "walltime_cutoff"
-                            : degenerate
-                                ? "degenerate"
-                                : !safetyCut && toolError === undefined && truncatedEmpty && !dynamicCut && !staticReasoningCut
-                                    ? "length_empty"
-                                    : undefined;
+                        const errorKind = degenerate
+                            ? "degenerate"
+                            : !safetyCut && toolError === undefined && truncatedEmpty && !dynamicCut && !staticReasoningCut
+                                ? "length_empty"
+                                : undefined;
                         const finalMessage = {
                             ...partial,
                             content: finalContent.length > 0 ? finalContent : [{ type: "text", text: "" }],

package/dist/brain/openai.js

@@ -1,9 +1,8 @@
 import {} from "../internal/llm.js";
 import { repairTextToolCalls } from "./tool-call-repair.js";
 import { DEGENERATE_MESSAGE, trimDegenerateTail } from "./repetition.js";
-import { createRepetitionPoll, createWalltimeGate, parseStreamedToolArgs } from "./stream-shared.js";
+import { createRepetitionPoll, parseStreamedToolArgs } from "./stream-shared.js";
 import { mintFallbackToolCallId } from "./tool-call-id.js";
-import { WALLTIME_CUTOFF_MESSAGE } from "./walltime.js";
 import { emitBrainTelemetry } from "./status-sink.js";
 import { errorResultMediaNote, IMAGE_OMITTED_NO_VISION, imagesOmittedNoVisionNote, modelSupportsVision, sendableImages } from "./media-degrade.js";
 import { OPENAI_RESERVED, applyExtraBody, stripAuthHeaders } from "./request-params.js";
@@ -275,7 +274,6 @@ export function createOpenAIBrain(config = {}) {
             signal: options?.signal,
             config,
             httpLabel: "gateway",
-            callDeadlineMs: options?.callDeadlineMs,
             stallTimeouts: options?.stallTimeouts,
             buildRequest: () => {
                 const apiKey = options?.apiKey ?? config.apiKey;
@@ -330,7 +328,6 @@ export function createOpenAIBrain(config = {}) {
                 let degenerate = false;
                 let degenFace;
                 const rep = createRepetitionPoll();
-                const walltime = createWalltimeGate({ callDeadlineMs: options?.callDeadlineMs, isDegenerate: () => degenerate, cancel: () => ctrl.cancel() });
                 let lastDegenCheck = 0;
                 let lastDegenCheckReasoning = 0;
                 const detectRep = config.detectRepetition !== false;
@@ -410,10 +407,7 @@ export function createOpenAIBrain(config = {}) {
                     sealForRetry() {
                         closeThinking();
                     },
-                    onDeadline: () => walltime.onDeadline(),
                     onLine(rawLine) {
-                        if (walltime.shouldDropLine())
-                            return;
                         const line = rawLine.trim();
                         if (!line.startsWith("data:"))
                             return;
@@ -534,40 +528,35 @@ export function createOpenAIBrain(config = {}) {
                         const dynamicCut = truncatedEmpty && options?.maxTokensDynamic === true;
                         const safetyCut = finishReason === "content_filter";
                         const refused = accumRefusal.trim().length > 0;
-                        const emptyNoFinish = noUsableContent && finishReason == null && !degenerate && !walltime.cut;
+                        const emptyNoFinish = noUsableContent && finishReason == null && !degenerate;
                         const maxTokensNote = sentMaxTokens !== undefined
                             ? `effective max_tokens=${sentMaxTokens} (from ${sentMaxTokensLane === "options" ? "options.maxTokens" : "model.maxTokens"})`
                             : "max_tokens not set (neither options.maxTokens nor model.maxTokens — the provider's default cap applied)";
-                        const errorMessage = walltime.cut
-                            ? WALLTIME_CUTOFF_MESSAGE
-                            : degenerate
-                                ? DEGENERATE_MESSAGE
-                                : safetyCut
-                                    ? `[refusal] response cut by the provider content filter (finish_reason="content_filter") — the output was censored/truncated for policy and is unreliable`
-                                    : refused
-                                        ? `[refusal] response refused by the model for policy reasons: ${accumRefusal.trim()}`
-                                        : (toolError ??
-                                            (truncatedEmpty && !dynamicCut
-                                                ? accumReasoning.trim()
-                                                    ? `[length_empty] response truncated at max_tokens (finish_reason="length", ${maxTokensNote}) with no answer text — the model spent the whole budget on reasoning (${accumReasoning.length} reasoning chars). Raise max_tokens, or disable thinking for this task.`
-                                                    : `[length_empty] response truncated at max_tokens (finish_reason="length", ${maxTokensNote}) with no answer text — raise max_tokens`
-                                                : emptyNoFinish
-                                                    ? `[stream_torn] model stream ended with no content and no finish_reason — the response was lost (a torn stream or an in-band provider error frame)${malformedFrames > 0 ? ` — ${malformedFrames} unparseable data frame(s) were dropped on this stream (first sample: ${JSON.stringify(malformedSample)})` : ""}`
-                                                    : undefined));
-                        const errored = walltime.cut ||
-                            degenerate ||
+                        const errorMessage = degenerate
+                            ? DEGENERATE_MESSAGE
+                            : safetyCut
+                                ? `[refusal] response cut by the provider content filter (finish_reason="content_filter") — the output was censored/truncated for policy and is unreliable`
+                                : refused
+                                    ? `[refusal] response refused by the model for policy reasons: ${accumRefusal.trim()}`
+                                    : (toolError ??
+                                        (truncatedEmpty && !dynamicCut
+                                            ? accumReasoning.trim()
+                                                ? `[length_empty] response truncated at max_tokens (finish_reason="length", ${maxTokensNote}) with no answer text — the model spent the whole budget on reasoning (${accumReasoning.length} reasoning chars). Raise max_tokens, or disable thinking for this task.`
+                                                : `[length_empty] response truncated at max_tokens (finish_reason="length", ${maxTokensNote}) with no answer text — raise max_tokens`
+                                            : emptyNoFinish
+                                                ? `[stream_torn] model stream ended with no content and no finish_reason — the response was lost (a torn stream or an in-band provider error frame)${malformedFrames > 0 ? ` — ${malformedFrames} unparseable data frame(s) were dropped on this stream (first sample: ${JSON.stringify(malformedSample)})` : ""}`
+                                                : undefined));
+                        const errored = degenerate ||
                             safetyCut ||
                             refused ||
                             emptyNoFinish ||
                             (toolError !== undefined && noUsableContent) ||
                             (truncatedEmpty && !dynamicCut);
-                        const errorKind = walltime.cut
-                            ? "walltime_cutoff"
-                            : degenerate
-                                ? "degenerate"
-                                : !safetyCut && !refused && toolError === undefined && truncatedEmpty && !dynamicCut
-                                    ? "length_empty"
-                                    : undefined;
+                        const errorKind = degenerate
+                            ? "degenerate"
+                            : !safetyCut && !refused && toolError === undefined && truncatedEmpty && !dynamicCut
+                                ? "length_empty"
+                                : undefined;
                         const finalMessage = {
                             ...partial,
                             content: finalContent.length > 0 ? finalContent : [{ type: "text", text: "" }],

package/dist/brain/stream-engine.d.ts

@@ -18,7 +18,6 @@ export interface StreamControls {
 }
 export interface StreamParser {
     onLine(line: string): void;
-    onDeadline?(): boolean;
     finalize(): void;
     snapshot(): {
         hasSubstantiveText: boolean;
@@ -35,7 +34,6 @@ export declare function runStreamingBrain(args: {
     httpLabel: string;
     buildRequest: () => SSERequest;
     makeParser: (ctrl: StreamControls) => StreamParser;
-    callDeadlineMs?: number;
     stallTimeouts?: {
         connectMs?: number;
         firstTokenMs?: number;