@sema-agent/core 5.57.0 → 5.59.0

package/dist/bench/metrics.d.ts

@@ -71,7 +71,7 @@ export interface DeliveryDecision {
 }
 /**
  * design/95 §3.2 — the lifecycle status of a run, so INFRA noise never poisons the mode signal
- * (codex BLOCKER B4 / design/89 §3.3 Beatsep red line). The Beatsep task#2b 9% pass-rate was infra
+ * (design/89 §3.3 Beatsep red line). The Beatsep task#2b 9% pass-rate was infra
  * death (OOM / K8S passthrough / nested-root), NOT a mode signal — feeding such runs into the
  * numerator turns them into spurious DELIVERED-WRONG / zero datapoints and reproduces the very noise
  * the design claims to have isolated. `buildReport` SCORES only `scored`; `infra-failed` / `excluded`
@@ -94,7 +94,7 @@ export interface RunRecord {
     /** Repeat/seed index within the cell (design/89 §3.4 N-repeat). */
     seed?: number | string;
     /**
-     * design/95 §3.2 / codex B4 — lifecycle status. Only "scored" runs reach the numerator / cost /
+     * design/95 §3.2 — lifecycle status. Only "scored" runs reach the numerator / cost /
      * Pareto. Absent ⇒ "scored" (back-compat). See {@link RunStatus}.
      */
     runStatus?: RunStatus;
@@ -313,7 +313,7 @@ export interface ArmCell {
     /** Avoided-loss rate = CORRECTLY-WITHHELD / n. The supervisor's §2.2.1 value, single-listed. */
     correctlyWithheldRate: number;
     /**
-     * codex MINOR / design/95 M10 — withhold treated as a binary detector of would-be-wrong delivery
+     * design/95 M10 — withhold treated as a binary detector of would-be-wrong delivery
      * (positive = withheld; ground truth = would-be-wrong, established by the §2.4 counterfactual). Reported
      * as precision/recall — NOT just the raw correctly-withheld count — so an arm cannot look good by
      * withholding indiscriminately (high count, low precision) or by rarely withholding (high precision,
@@ -420,10 +420,10 @@ export declare function riskTransferDisclosure(cells: ArmCell[]): {
     };
 };
 /**
- * codex B2 — a paired-binary comparison of two arms on the SAME tasks/seeds. The two arms' truly-
+ * a paired-binary comparison of two arms on the SAME tasks/seeds. The two arms' truly-
  * correct flags are paired row-by-row (paired-seed, design/89 §3.4). For binary paired data the
  * RIGHT tests are McNemar (discordant pairs) and a paired-bootstrap difference interval — NOT two
- * independent proportions with "CI non-overlap" (that ignores the pairing and is under-powered, codex B2).
+ * independent proportions with "CI non-overlap" (that ignores the pairing and is under-powered).
  */
 export interface PairedBinaryComparison {
     /** Arm A (e.g. "sup") vs arm B (e.g. "solo"). pDiff = P(A) − P(B). */
@@ -440,12 +440,12 @@ export interface PairedBinaryComparison {
      *  b+c large). */
     mcnemarP: number;
     /** Bootstrap percentile 95% CI for the paired difference (lo, hi) — a percentile CI on the seeded
-     *  paired-bootstrap distribution, NOT Newcombe's analytic interval (codex MAJOR-B). Crosses 0 ⇒ not significant. */
+     *  paired-bootstrap distribution, NOT Newcombe's analytic interval. Crosses 0 ⇒ not significant. */
     ci95: [number, number];
     /** TRUE iff the difference is statistically significant at α=0.05 (CI excludes 0 AND McNemar p<0.05). */
     significant: boolean;
     /**
-     * codex B2 — pre-registered Minimum Detectable Effect at the observed nPairs (the difference this
+     * pre-registered Minimum Detectable Effect at the observed nPairs (the difference this
      * comparison COULD have detected at 80% power). When |pDiff| is below this AND not significant, the
      * verdict is "not powered" — NOT "no difference". This is the field that stops "N≥15 is an assertion".
      */
@@ -456,8 +456,8 @@ export interface PairedBinaryComparison {
     verdict: "A-better" | "B-better" | "no-detectable-difference" | "not-powered";
 }
 /**
- * codex B2 — compare two arms' delivered truly-correct as PAIRED binary. Pairs runs by the COMPOSITE
- * `(taskId, seed)` key (codex MAJOR-A: a bare `seed` collides across tasks — the same repeat index
+ * compare two arms' delivered truly-correct as PAIRED binary. Pairs runs by the COMPOSITE
+ * `(taskId, seed)` key (a bare `seed` collides across tasks — the same repeat index
  * recurs per task — so cross-task input would overwrite pairs and contaminate the McNemar sample);
  * only rows where BOTH arms delivered (not withheld, both scored) form a pair — a
  * withheld run has no delivered binary to pair (it is scored in the withhold/avoided-loss axis, not
@@ -469,7 +469,7 @@ export declare function pairedBinaryCompare(scoredRuns: RunRecord[], armA: Arm,
     bootstrapIters?: number;
 }): PairedBinaryComparison;
 /**
- * codex B3 — the pre-registered C2 exchange rate(s): how many human-review SECONDS we are willing to
+ * the pre-registered C2 exchange rate(s): how many human-review SECONDS we are willing to
  * pay to buy one unit of supervisor value. Without these, "higher P / more withholds ⇒ worth it" is
  * unfalsifiable (the "helpful but too expensive" counter-thesis cannot be observed). Declared BEFORE
  * the run (design/95 §9), not fit after.
@@ -482,20 +482,20 @@ export interface C2Thresholds {
 }
 export interface ParetoVerdict {
     arm: Arm;
-    /** Non-dominated on (C1, C2.sec, P) — necessary condition to be considered at all (codex B3). */
+    /** Non-dominated on (C1, C2.sec, P) — necessary condition to be considered at all. */
     nonDominated: boolean;
     /** The pre-registered exchange-rate check: does the arm's extra C2 buy enough avoided-loss / saved
      *  decisions to clear the declared threshold? Undefined when the arm has no extra C2 over baseline. */
     clearsC2Threshold?: boolean;
     /**
-     * codex B3 — the label. An arm may ONLY be "worth-it" when it is non-dominated AND clears the C2
+     * the label. An arm may ONLY be "worth-it" when it is non-dominated AND clears the C2
      * threshold. A non-dominated but threshold-failing arm is "quality-tradeoff" (higher quality, but the
      * cost is not bought back) — NEVER "worth-it". A dominated arm is "dominated".
      */
     label: "worth-it" | "quality-tradeoff" | "dominated";
 }
 /**
- * codex B3 — classify each arm against the SUP/TEAM-vs-baseline value question with a pre-registered
+ * classify each arm against the SUP/TEAM-vs-baseline value question with a pre-registered
  * exchange rate. `baselineArm` is the cost floor to compare extra C2 against (default "solo").
  * `avoidedWrong` / `savedDecisions` per arm come from the cells / campaign report.
  *
@@ -518,31 +518,31 @@ export declare function paretoValueVerdict(cells: ArmCell[], thresholds: C2Thres
  * design/95 §7.1 — V1 (免重复劳动): campaign-level human-decision delta. The ONLY authoritative V1
  * mechanism (design/95 reconcile, MAJOR#2): cross-task reuse of a decided strategy. Semantics:
  *   - `soloHumanDecisions` is the V1 BASELINE — the operator's actual up-front decision count when
- *     running solo across the campaign (codex M7: NOT 0, NOT synthetic — measured from real operator
+ *     running solo across the campaign (NOT 0, NOT synthetic — measured from real operator
  *     prep across runs). `0` is only valid when the campaign genuinely needed no human decision.
  *   - `repeatedDecisionsSaved = soloHumanDecisions − supHumanDecisions` (may be negative = SUP cost
  *     MORE human decisions; reported honestly, not floored).
  */
 export interface CampaignV1Saved {
     decisionKind: string;
-    /** V1 baseline (codex M7): operator's measured up-front decisions when running SOLO. Provenance MUST
+    /** V1 baseline: operator's measured up-front decisions when running SOLO. Provenance MUST
      *  be recorded in the stamp (real prep, not 0/synthetic). */
     soloHumanDecisions: number;
     /** SUP arm's actual human-review decision count for this decision kind. */
     supHumanDecisions: number;
     /** = soloHumanDecisions − supHumanDecisions. Positive ⇒ SUP saved repeated labour (V1 evidence). */
     repeatedDecisionsSaved: number;
-    /** Whether the baseline is a real measurement vs absent (codex M7 honesty gate). */
+    /** Whether the baseline is a real measurement vs absent (honesty gate). */
     baselineProvenance: "measured-operator-baseline" | "absent-not-claimable";
 }
 /**
- * design/95 §7.2 — V2 (蓝图清晰度) THREE-arm ablation (codex BLOCKER#1 + MAJOR#3 / M8). All three are
+ * design/95 §7.2 — V2 (蓝图清晰度) THREE-arm ablation. All three are
  * paired on the SAME seeds (paired===true ⇒ the three pX arrays/values are over the same n seeds).
  * INVARIANT: `n` is the paired-seed count common to all three sub-arms; the three deltas are computed
  * over those n pairs.
  *   - pSoloRaw            — only the original brief (no blueprint baseline).
  *   - pSoloSupGenerated   — the SUP-arm-GENERATED blueprint (attributable to supervisor; gen cost charged).
- *   - pSoloExperimenterIdeal — the experimenter "ideal" blueprint. codex M8: this is a CONSTRAINED upper
+ *   - pSoloExperimenterIdeal — the experimenter "ideal" blueprint. this is a CONSTRAINED upper
  *     bound on "better-prompt help", NOT "supervisor's mechanistic ceiling", UNLESS the ideal blueprint
  *     was generated under the {@link IdealBlueprintConstraints} (run-time, visible-brief-only, no oracle
  *     access, same info budget, leak-reviewed). `idealConstraintsSatisfied` records which it is.
@@ -557,16 +557,16 @@ export interface BlueprintAblationTriple {
     pSoloExperimenterIdeal: number;
     /** = pSoloSupGenerated − pSoloRaw (supervisor blueprint's real net value). */
     deltaSupVsRaw: number;
-    /** = pSoloExperimenterIdeal − pSoloRaw. LABELLED per `idealConstraintsSatisfied` (codex M8). */
+    /** = pSoloExperimenterIdeal − pSoloRaw. LABELLED per `idealConstraintsSatisfied`. */
     deltaIdealVsRaw: number;
-    /** codex M8 — when false, deltaIdealVsRaw is "manual-prompt upper bound", NOT "supervisor potential". */
+    /** when false, deltaIdealVsRaw is "manual-prompt upper bound", NOT "supervisor potential". */
     idealConstraintsSatisfied: boolean;
-    /** codex M3 — blueprint generation cost charged to the sup-generated arm (budget-match red line). */
+    /** blueprint generation cost charged to the sup-generated arm (budget-match red line). */
     blueprintGenC1MicroUsd: number;
     blueprintGenC2Count: number;
     blueprintGenC2Sec: number;
 }
-/** codex M8 — the constraints under which an "ideal" blueprint may be read as a supervisor-mechanism
+/** the constraints under which an "ideal" blueprint may be read as a supervisor-mechanism
  *  upper bound rather than a generic "better prompt helps" result. Recorded per ablation. */
 export interface IdealBlueprintConstraints {
     generatedBeforeRun: boolean;
@@ -575,10 +575,10 @@ export interface IdealBlueprintConstraints {
     sameInfoBudget: boolean;
     solutionLeakReviewed: boolean;
 }
-/** True iff ALL ideal-blueprint constraints hold (codex M8 gate). */
+/** True iff ALL ideal-blueprint constraints hold. */
 export declare function idealConstraintsSatisfied(c: IdealBlueprintConstraints): boolean;
 /**
- * design/95 §7.3 — V3 (无关性隔离) contamination probe (canary, mechanical, zero LLM-judge). codex M9:
+ * design/95 §7.3 — V3 (无关性隔离) contamination probe (canary, mechanical, zero LLM-judge).
  * the isolated-vs-shared P delta CONFOUNDS execution order / context size / worker count / budget; the
  * PRIMARY signal must be per-subgoal oracle failure + explicit wrong-use of a sibling artifact (the
  * canary leak), with the P delta kept as a DIAGNOSTIC only.
@@ -590,12 +590,12 @@ export interface IsolationContamination {
     isolatedSubgoalP: number[];
     /** Per-factor P under shared context. length === factorCount. */
     sharedSubgoalP: number[];
-    /** PRIMARY mechanical signal (codex M9): cross-factor canary leak rate (explicit token bleed). */
+    /** PRIMARY mechanical signal: cross-factor canary leak rate (explicit token bleed). */
     canaryLeakRate: number;
-    /** PRIMARY mechanical signal (codex M9): count of factors that failed their oracle AND demonstrably
+    /** PRIMARY mechanical signal: count of factors that failed their oracle AND demonstrably
      *  used a sibling factor's artifact (the controlled-ablation harm signal). */
     wrongSiblingArtifactUses: number;
-    /** DIAGNOSTIC ONLY (codex M9 confound): mean(isolated P) − mean(shared P). Not the primary signal. */
+    /** DIAGNOSTIC ONLY (confound): mean(isolated P) − mean(shared P). Not the primary signal. */
     contaminationRateDiag: number;
     /** Extra C1 the isolation cost (multi-worker / repeated context load). */
     isolationOverheadC1: number;
@@ -614,7 +614,7 @@ export interface CampaignReport {
     implementedAxes: ImplementedAxes;
 }
 /**
- * design/95 §2 / codex B1 — one heterogeneity CELL: a `(suiteVersion, taskId, archetype, valueDimension,
+ * design/95 §2 — one heterogeneity CELL: a `(suiteVersion, taskId, archetype, valueDimension,
  * arm)` group with its folded {@link ArmCell}. The decision surface is the per-cell vector — NOT the
  * arm mean (the Simpson's-paradox guard). `groupKey` is the stable join key.
  */
@@ -648,7 +648,7 @@ export interface ComparisonCell {
     pareto: ParetoResult;
 }
 /**
- * codex MINOR11 / council Q1 — capability metadata: which of the four §7/Pareto value axes are actually
+ * council Q1 — capability metadata: which of the four §7/Pareto value axes are actually
  * IMPLEMENTED in this report. Consumers must read this before claiming a V1/V2/V3 result; an unset axis
  * is NOT a null result, it is "not measured here". (The T1 structures above are types + pure helpers;
  * the report-level V1/V2/V3 fields are populated by the harness, not synthesized in `buildReport`.)
@@ -660,7 +660,7 @@ export interface ImplementedAxes {
     v3: boolean;
 }
 export interface ValueJudgmentReport {
-    /** codex B1 — per-arm cells grouped by full heterogeneity key (incl. arm). Single-arm; use for cell
+    /** per-arm cells grouped by full heterogeneity key (incl. arm). Single-arm; use for cell
      *  inspection / `pairedBinaryCompare` inputs. NOT directly a cross-arm verdict surface (each is one arm). */
     groupedCells: GroupedCell[];
     /**
@@ -671,7 +671,7 @@ export interface ValueJudgmentReport {
      */
     comparisons: ComparisonCell[];
     /**
-     * codex B1 — suite-level rollup: per-arm cells folded over ALL scored runs. A WEIGHTED summary for
+     * suite-level rollup: per-arm cells folded over ALL scored runs. A WEIGHTED summary for
      * dashboards ONLY — NOT a decision surface (folding heterogeneous tasks can flip a verdict, Simpson).
      * Consumers MUST decide on `comparisons`; this is convenience aggregation.
      */
@@ -687,7 +687,7 @@ export interface ValueJudgmentReport {
     /** design/89 §3.4: cells with n<8 are directional-only, not verdict-grade. Flagged, not dropped. */
     directionalOnly: Arm[];
     /**
-     * codex B4 — the exclusion ledger: runs dropped from scoring (infra-failed / excluded) with reasons.
+     * the exclusion ledger: runs dropped from scoring (infra-failed / excluded) with reasons.
      * A regression-grade invariant: these never reach `cells` / `pareto` (no infra noise in the verdict).
      */
     excluded: Array<{
@@ -696,16 +696,16 @@ export interface ValueJudgmentReport {
         seed?: number | string;
         reason: RunStatus;
     }>;
-    /** codex MINOR11 — which value axes this report actually measured. */
+    /** which value axes this report actually measured. */
     implementedAxes: ImplementedAxes;
 }
 /**
- * Top-level (codex B1+B4): SCORE only `scored` runs (B4 — drop infra-failed/excluded with reasons),
+ * Top-level: SCORE only `scored` runs (B4 — drop infra-failed/excluded with reasons),
  * group by `(suiteVersion, taskId, archetype, valueDimension, arm)` (B1 — the per-cell decision surface,
  * Simpson's-paradox guard), fold each group, AND provide a suite-level per-arm rollup for dashboards
  * (explicitly NOT a decision surface). Deterministic; no model.
  *
- * `opts.implementedAxes` lets the harness declare which §7 axes it populated (codex MINOR11); default
+ * `opts.implementedAxes` lets the harness declare which §7 axes it populated; default
  * = only Pareto (the V1/V2/V3 structures are T1 net-new, not synthesized here).
  */
 export declare function buildReport(runs: RunRecord[], opts?: {

package/dist/brain/degrading.d.ts

@@ -75,7 +75,7 @@ export declare function readDegradation(msg: {
     diagnostics?: AssistantMessageDiagnostic[];
 } | undefined): DegradationInfo | undefined;
 /**
- * design/131 (codex 131 审 A) — resilience-standdown pass-through: forward one brain's stream
+ * design/131 — resilience-standdown pass-through: forward one brain's stream
  * verbatim, but keep the decorator discipline that a REJECTING inner brain (throw instead of an
  * error event) still leaves a terminal on the returned stream — otherwise `result()` never
  * resolves and the harness hangs (the same class as the degradation-brain blocker).

package/dist/brain/errors.d.ts

@@ -12,7 +12,7 @@ export declare const ZERO_USAGE: Usage;
 /** The single source of truth for brain error codes — the type, the `[code]` parser, and the prefix
  *  stripper all derive from this one list, so adding a code can't silently diverge across call sites.
  *
- *  The last three are IN-BAND terminal-output codes ([1086]③): stamped on errored assistant MESSAGES
+ *  The last three are IN-BAND terminal-output codes: stamped on errored assistant MESSAGES
  *  (never thrown as BrainError) so `TaskResult.errorCode` covers the error family end-to-end and
  *  downstream consumers can bucket without string-matching. Deliberately behavior-neutral everywhere
  *  else: the circuit breaker's default countCodes excludes them (model-output problems, not provider
@@ -27,7 +27,7 @@ export type BrainErrorCode = (typeof BRAIN_ERROR_CODES)[number];
  *  tool-errors.ts must classify every member — a code added here but missed there folds to "unknown"). */
 export declare const ALL_BRAIN_ERROR_CODES: readonly BrainErrorCode[];
 /** The in-band terminal-output subset (see the list doc above). The circuit breaker collapses these
- *  to the legacy `"http"` bucket before counting (codex [1086]③ 审): its taxonomy is TRANSPORT-level,
+ *  to the legacy `"http"` bucket before counting: its taxonomy is TRANSPORT-level,
  *  and before these codes existed the same failures reached it as the unprefixed→"http" fallback — a
  *  custom `countCodes: ["http"]` policy must keep counting them exactly as it always did. */
 export declare const IN_BAND_OUTPUT_CODES: ReadonlySet<BrainErrorCode>;
@@ -65,7 +65,7 @@ export declare function stripErrorCodePrefix(errorMessage: string): string;
  */
 export type ConnectFailureClass = "deterministic" | "indeterminate";
 export declare function classifyConnectFailure(e: unknown): ConnectFailureClass;
-/** RB-386③ ([2090]) — render a transport-level error WITHOUT folding away its cause chain. undici's
+/** RB-386③ — render a transport-level error WITHOUT folding away its cause chain. undici's
  *  `fetch` rejects with `TypeError("fetch failed", { cause })` where the CAUSE carries the actually
  *  actionable fact (`connect ECONNREFUSED …` / `getaddrinfo ENOTFOUND …` / `SocketError: other side
  *  closed` with a `code`); `String(e)` / `e.message` alone collapsed ALL THREE network-failure shapes

package/dist/brain/reasoning.d.ts

@@ -92,7 +92,7 @@ export interface ReasoningResolution {
  *  the cap-wins predicate below and the brain's budget-window math read ONE constant. */
 export declare const MIN_THINKING_TOKENS = 1024;
 /**
- * The anthropic BUDGET path's cap-wins arm (#346, single-sourced; design/119 #2 review codex H4):
+ * The anthropic BUDGET path's cap-wins arm (#346, single-sourced; design/119 #2 review):
  * a HARD per-request output cap (an engine-imposed override or the caller's explicit
  * `options.maxTokens` — the two lanes the brain refuses to raise) too small to host a legal thinking
  * budget (≥ {@link MIN_THINKING_TOKENS}) plus answer room means the CAP WINS and thinking is skipped
@@ -276,7 +276,7 @@ export interface ReasoningProfileFlags {
 }
 /**
  * design/96 §D (S4) — map a named {@link ReasoningTier} to its engine {@link ReasoningProfileFlags} (default
- * mapping). G2 (clay): the intensity→capability mapping is PROFILE-layer data — core gives this primitive +
+ * mapping). G2: the intensity→capability mapping is PROFILE-layer data — core gives this primitive +
  * a sane default; a profile/scenario layer calls it (and may override the table) to set a task's `thinking` +
  * prompt flags. Core never forces it. Returns a fresh COPY (callers may mutate).
  */

package/dist/brain/repetition.d.ts

@@ -36,7 +36,7 @@ export declare const DEGENERATE_MESSAGE = "degenerate repetition detected \u2014
  * One detector hit — either the event that CUT a stream (`degenerate`) or a repetition that landed in
  * a detection window but was SPARED by a 2e1c161 structural allowance (code-line shape / divider run).
  * Spared events are the RSI raw material for judging whether the allowance is too wide or too narrow
- * (clay 2026-07-10: the detector must report counts AND the detected segment, not just a boolean).
+ * (2026-07-10: the detector must report counts AND the detected segment, not just a boolean).
  */
 export interface RepetitionEvent {
     /** Which rule's window the repetition landed in: rule 1 (`char-run`) or rule 2 (`unit-loop`). */

package/dist/brain/route-adjudicator.d.ts

@@ -174,7 +174,13 @@ export declare function adjudicateDerivedRoute(args: {
  */
 export declare function sameRouteIdentity(a: Pick<Model, "id" | "baseUrl">, b: Pick<Model, "id" | "baseUrl">): boolean;
 /** The `route.fallback_to_primary` announcement — one shared shape for every derived-leg seat, so
- *  a consumer can rely on `detail.{seat, from, to, cause}` regardless of which seat fell back. */
+ *  a consumer can rely on `detail.{seat, from, to, cause}` regardless of which seat fell back.
+ *
+ *  `sessionId` (#433, additive): the swap happened while resolving ONE run's seats, so the run is
+ *  what an operator correlates the line against ("which task ran on the fallback model"). The
+ *  AUDIENCE stays `"operator"` — the fix is a credential/catalog fix, and no end user can act on it;
+ *  attribution is a routing key, not an entitlement. Omitted when a caller has no session in hand,
+ *  never fabricated. */
 export declare function fallbackToPrimaryNotice(args: {
     seat: string;
     from: string;
@@ -182,6 +188,7 @@ export declare function fallbackToPrimaryNotice(args: {
     verdict: Extract<RouteAdjudication, {
         ok: false;
     }>;
+    sessionId?: string;
 }): EngineNotice;
 /** The auth carriers of a header bag, canonicalized (case-fold name → value, sorted) — the
  *  Model-visible credential fingerprint used by catalog-change disclosure (a same-name entry whose

package/dist/brain/route-adjudicator.js

@@ -173,7 +173,14 @@ export function fallbackToPrimaryNotice(args) {
         code: "route.fallback_to_primary",
         message: `derived model "${args.from}" (${args.seat}) is not routable (${args.verdict.code}) — ` +
             `falling back to the primary model "${args.to}". ${args.verdict.detail.fixHint}`,
-        detail: { seat: args.seat, from: args.from, to: args.to, cause: args.verdict.code, fixHint: args.verdict.detail.fixHint },
+        detail: {
+            seat: args.seat,
+            from: args.from,
+            to: args.to,
+            cause: args.verdict.code,
+            fixHint: args.verdict.detail.fixHint,
+            ...(args.sessionId !== undefined ? { sessionId: args.sessionId } : {}),
+        },
     };
 }
 export function authCarrierFingerprint(headers) {

package/dist/brain/status-sink.d.ts

@@ -18,7 +18,7 @@ export declare function runWithStatusSink<T>(emit: (s: BrainStatus) => void, fn:
  */
 export declare function emitBrainStatus(status: BrainStatus): void;
 /**
- * Silent-fallback telemetry (service [398] C1/C4/C5/C6) — the brain→runner TELEMETRY channel, a
+ * Silent-fallback telemetry (C1/C4/C5/C6) — the brain→runner TELEMETRY channel, a
  * SECOND ALS sink parallel to the status one. Statuses are user-facing liveness frames (wire
  * `status` events); telemetry is operator-facing fallback accounting the Runner bridges to trace
  * events (`brain.failover` / `breaker.transition` / `brain.retry` / `vision.placeholder`). Kept
@@ -52,7 +52,7 @@ export type BrainTelemetry = {
 }
 /** PDF read: a `document` content block was substituted with a text placeholder because the serving
  *  API/model cannot take native document input (openai brain always; anthropic brain when the model
- *  declares no vision). Silent quality loss → frame (same [398] C4 discipline as vision_placeholder). */
+ *  declares no vision). Silent quality loss → frame (same C4 discipline as vision_placeholder). */
  | {
     kind: "document_placeholder";
     count: number;

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

@@ -35,7 +35,7 @@ export interface RepetitionPoll {
     readonly spared: RepetitionEvent[];
 }
 /**
- * Repetition telemetry (clay 2026-07-10): the cut event + allowance-spared hits for THIS stream.
+ * Repetition telemetry (2026-07-10): the cut event + allowance-spared hits for THIS stream.
  * Polls re-inspect the same growing text every ~64 chars — dedupe spared by rule+segment so one
  * boot block doesn't log once per poll. Attached to the final message as `repetition`.
  */

package/dist/config/catalog.d.ts

@@ -1,5 +1,5 @@
 /**
- * Config catalog (clay ruling 2026-07-19) — the SINGLE machine-readable source for the engineering
+ * Config catalog (ruled 2026-07-19) — the SINGLE machine-readable source for the engineering
  * configuration scale across every client (server / cli / TB harness / future shells).
  *
  * Why this exists: the same knob was found carrying FIVE different values across layers (maxTurns:
@@ -12,8 +12,8 @@
  *      nothing) or overrides it and DECLARES the override (`TaskSpec.configOverrides`, with a
  *      reason). Declared overrides ride the `config.assembled` trace, so every run self-reports its
  *      true knobs and their provenance — "who set 80" becomes a trace read, not archaeology.
- *   3. **Announced change discipline** (clay): the recommendations are a living baseline — any repo
- *      changing one (or overriding persistently) posts the change to the shared blackboard so the
+ *   3. **Announced change discipline**: the recommendations are a living baseline — any repo
+ *      changing one (or overriding persistently) posts the change to the shared config record so the
  *      scale evolves deliberately, together.
  *
  * v1 scope: the cross-client TASK-LAUNCH knobs (the chaos surface). Internal engine tuning
@@ -62,12 +62,12 @@ export interface EffectiveConfigField {
     value: number | boolean | string | null;
     provenance: ConfigProvenance;
 }
-/** Declaration bounds (codex): host strings ride the trace stream — bounded and sanitized, never
+/** Declaration bounds: host strings ride the trace stream — bounded and sanitized, never
  *  free-form. Beyond `MAX_OVERRIDE_DECLARATIONS` entries are dropped (count-capped, first-N kept). */
 export declare const MAX_OVERRIDE_DECLARATIONS = 32;
 /**
  * Build the effective task-launch config + provenance from a spec's launch fields. Pure. The
- * manifest must mirror ENGINE truth, not the spec's surface (codex): `maxOutputTokens` falls back
+ * manifest must mirror ENGINE truth, not the spec's surface: `maxOutputTokens` falls back
  * to the model's own cap (provenance `derived`) before "provider default".
  */
 export declare function resolveEffectiveConfig(spec: {

package/dist/core/arg-summary.d.ts

@@ -3,11 +3,11 @@
  * display (the MF-W workflow monitor's `Read(path)` / `Bash(grep …)` activity line; the MF-24 permission-denial
  * ledger's denied-tool input). Shared so both the activity sink and the denial record scrub identically — a
  * single audited boundary. NOT `redactHostLeaks` (that is for untrusted-LLM egress and over-redacts the very
- * path the trusted owner-scoped monitor wants to show). codex-reviewed (3 BLOCKER + 5 MAJOR closed).
+ * path the trusted owner-scoped monitor wants to show). Reviewed: 3 blocker + 5 major findings closed.
  */
 /** Known credential SHAPES + `secretKey=value` — redacted from the FINAL abbreviated string, the single scrub
- *  point before any store/SSE fan-out (codex F-5/F-6/F-8/F-9/F-10). Conservative (real credential formats). */
-/** [1608] every prefix is LEFT-ANCHORED with \`(?<![A-Za-z0-9])\` — without it, ordinary words whose
+ *  point before any store/SSE fan-out. Conservative (real credential formats). */
+/** every prefix is LEFT-ANCHORED with \`(?<![A-Za-z0-9])\` — without it, ordinary words whose
  *  TAIL spells a prefix were silently redacted (NETWORK-attached → NETWO[redacted], RTMARK-100,
  *  task_categories, monkey=abcd1234, MonkeyJumping…) while triggering unpredictably on suffix
  *  length. \`_\` is deliberately NOT in the anchor class: \`api_key=\`, \`STRIPE=sk_live_…\`-style glued
@@ -82,7 +82,7 @@ export declare const SECRET_PASSES: readonly RedactionPass[];
 export declare function scrubSecrets(s: string, report?: RedactionReport): string;
 /** A SHORT, SECRET-SCRUBBED primary-arg summary. Abbreviation REDUCES the surface; a final {@link scrubSecrets}
  *  pass is the leak boundary (the value is PERSISTED + SSE-forwarded and may be read by a platform operator in a
- *  multi-tenant deployment — codex F-10). Rules (codex-reviewed):
+ *  multi-tenant deployment). Rules:
  *   • command → the command NAME (skip leading `VAR=val` env-prefixes; bail to `undefined` on a quoted assignment
  *     or a lone assignment; basename a path-form command; `+ …` if it had args).
  *   • file path → BASENAME only (no host directory layout).

package/dist/core/ask-class.d.ts

@@ -2,7 +2,7 @@
  * F-012 L2 — the ENGINE-INTERNAL ask-class envelope.
  *
  * An `ask` verdict's SEMANTIC CLASS decides whether a deployment-declared sandbox may auto-admit it
- * (clay's ruling: a declared sandbox admits everything inside it — the boundary crossings park to a
+ * (the ruling: a declared sandbox admits everything inside it — the boundary crossings park to a
  * human, escape prevention belongs to the sandbox, not to engine re-policing). The class deliberately
  * lives OUTSIDE the public `PermissionResult`: a decision object is caller-supplied data, and a
  * self-reported "this ask is sandbox-local" field would be the same trust hole as the
@@ -27,7 +27,7 @@ export type AskClass = "sandbox_local" | "external_authority" | "hook" | "mandat
 /**
  * Brand a built-in policy instance with its ask class (engine factories only — not exported from the
  * package). The `check` binding is LOCKED as a non-writable, non-configurable DATA property before the
- * policy is returned (codex falsification-style round 4, HIGH): reading `policy.check` and INVOKING it must
+ * policy is returned (falsification-style round 4, HIGH): reading `policy.check` and INVOKING it must
  * name the same callable, atomically. A plain `===` re-read was TOCTOU-able by an accessor — a getter
  * could return the branded function to `policyAskClassOf` and a deployment ask to the gate's
  * invocation. Locking the property makes the callable un-swappable and un-redefinable, so the class

package/dist/core/ask-question.d.ts

@@ -23,7 +23,7 @@ export declare const ASK_USER_QUESTION_TOOL_NAME = "AskUserQuestion";
  * injection vector. So a `selected` value is echoed as a chosen option ONLY if it is one of the offered option
  * labels (`selected ⊆ options` — the model's own text), and the operator's free-text (`note`, plus any off-list
  * input) is wrapped in an untrusted-DATA fence so it can never be read as instructions. Abort while waiting is
- * handled by the §9 interrupt path (the pending call is closed). Mounting ([884], CC non-interactive `isEnabled`
+ * handled by the §9 interrupt path (the pending call is closed). Mounting (CC non-interactive `isEnabled`
  * parity): only when a delivery face exists — a live `onQuestion`, or a durable park consumer (`checkpointStore`
  * + `durableApproval`/`forceDurableGate`), or an explicit `TaskSpec.interactiveTools: true` — see that knob.
  */

package/dist/core/auto-compaction.d.ts

@@ -103,7 +103,7 @@ export interface MaybeCompactOptions {
     /** The session to compact. Typed on the INTERFACE face (typecheck 车 2026-07-24): every member this
      *  module touches lives on `Session` (the optional `getPromptEpoch` degrades honestly via `?.()`). */
     session: Session;
-    /** [1068] ruling a — this run's provider-declared prompt sections (the epoch declaration axis):
+    /** This run's provider-declared prompt sections (the epoch declaration axis):
      *  the compaction-boundary epoch restatement/advance must hash the SAME artifact identity as the
      *  prepare-time pin, or every compaction under a declaration-carrying run would flap the pin. */
     epochDeclaredSections?: import("../prompt-assembly/epoch.js").EpochDeclaredSections;
@@ -113,12 +113,12 @@ export interface MaybeCompactOptions {
      *  runs ONLY after that append succeeded — a failed/blocked compaction must swap nothing. */
     centerAdoption?: {
         /** Absent together with `clear` ⇒ never happens; adoption carries the new digest, ROLLBACK
-         *  (explicit center disable, codex F1) carries `clear:true` and no digest. */
+         *  (explicit center disable) carries `clear:true` and no digest. */
         centerArtifactDigest?: string;
         sourceRevision?: string;
         clear?: true;
         declaredSections: import("../prompt-assembly/epoch.js").EpochDeclaredSections;
-        /** codex F2 — the post-swap prompt overhead in the trigger estimate's coordinate: used for THIS
+        /** the post-swap prompt overhead in the trigger estimate's coordinate: used for THIS
          *  compaction's post-measurement so PTL recovery/anti-thrash see the swapped prefix size. */
         overheadTokensAfter: number;
         apply: (committedArtifactDigest?: string) => void;
@@ -254,7 +254,7 @@ export interface MaybeCompactOptions {
         /** Per-file char cap (≈4 chars/token). Default 16_000 (~4k tokens). */
         maxCharsPerFile?: number;
         /**
-         * Blackboard 2026-07-03 (CC-parity selection set): the files the task has READ, most recent
+         * CC-parity selection set: the files the task has READ, most recent
          * first (CC 198's post-compact restore sorts readFileState by timestamp — the read set is wider
          * than the modified set and includes reference files the model consulted). When provided and
          * non-empty it WINS over `modifiedFilesByRecency`; absent/empty falls back to the modified set
@@ -272,7 +272,7 @@ export interface MaybeCompactOptions {
          */
         instructionSourcePaths?: ReadonlyArray<string>;
         /**
-         * RB-197②(独立复审 + codex 命中,已修): `recentlyReadFiles()` yields CANONICAL keys (resolveKey
+         * RB-197②(独立复审 + 命中,已修): `recentlyReadFiles` yields CANONICAL keys (resolveKey
          * output — the same coordinate `ReadFileState` is keyed in), but `instructionSourcePaths` and the
          * kept-tail scan inside `maybeCompact` (which reads raw tool-call arguments via
          * `extractFileOpsFromMessage`) do NOT share that coordinate — a model-typed relative path and its
@@ -287,7 +287,7 @@ export interface MaybeCompactOptions {
          */
         normalizePath?: (raw: string) => Promise<string>;
         /**
-         * RB-197②(codex 交叉复审命中,已修) — a kept-tail Read tool-CALL's mere presence does not prove the
+         * RB-197②(交叉复审命中,已修) — a kept-tail Read tool-CALL's mere presence does not prove the
          * file's real content is visible there: its toolRESULT may be a dedup stub ("unchanged since you
          * last read it…") whose ORIGINAL full transmission sits in the summarized prefix THIS compaction is
          * about to cut away — preserving read-state for such a path would leave both the kept tail and the
@@ -307,7 +307,7 @@ export interface MaybeCompactOptions {
      * `workingFileAttachments` deliberately — the invalidation is owed even when attachments are disabled
      * (an empty array then). Best-effort: a throwing callback must not fail the landed compaction.
      *
-     * RB-197②(codex 命中,已修;独立复审第二轮又命中一条同族缺口,已修) `preserveReadState`:
+     * RB-197②(命中,已修;独立复审第二轮又命中一条同族缺口,已修) `preserveReadState`:
      * canonical-coordinate paths (same shape as `normalizePath`'s output) whose read-state entry must
      * survive the compaction's clear even though they are NOT in `attachedComplete` — every file
      * `maybeCompact` deliberately skipped re-attaching, for EITHER exclusion reason (kept-tail-visible,
@@ -434,7 +434,7 @@ export interface MaybeCompactOptions {
      */
     postCompact?: Hooks["postCompact"];
     /**
-     * MF-18 修② ([496]③) — observe-only: fired when a summary call's serialized conversation INPUT was
+     * MF-18 修② — observe-only: fired when a summary call's serialized conversation INPUT was
      * truncated to fit the compaction model's context window (the ~300K-session shape that previously
      * guaranteed a provider prompt-too-long throw at every boundary). The caller bridges it to the
      * `compaction.input_truncated` trace frame (fidelity disclosure). Fired BEFORE the summary call, so
@@ -446,7 +446,7 @@ export interface MaybeCompactOptions {
         keptChars: number;
     }) => void;
     /**
-     * RB-402 ([2132]③c/[2133]) — the design/36 ToolSearch activation set LIVE at this pass (the
+     * RB-402 — the design/36 ToolSearch activation set LIVE at this pass (the
      * runner threads `[...prepared.activeTools]` on all three lanes: boundary, PTL recovery, finish).
      * Persisted into the landed compaction entry's details (`CompactionDetails.activeTools`) so the
      * NEXT task boundary's re-derivation — whose transcript source (ToolSearch toolCall parts) this
@@ -478,11 +478,11 @@ export declare function maybeCompact(opts: MaybeCompactOptions): Promise<{
      *  (auto/manual triggers only — never on "forced"). NOT a failure: the caller must not count it
      *  toward the compaction breaker; the next boundary consults the callback again. */
     blocked?: boolean;
-    /** MF-18 修① ([496]③): set (with `compacted:false`) when compaction is DISABLED by settings
+    /** MF-18 修①: set (with `compacted:false`) when compaction is DISABLED by settings
      *  (`enabled:false`) — the request can never be honored this run. Lets the caller resolve a manual
      *  /compact with the honest `"disabled"` outcome (and trace frame) instead of a structural noop. */
     disabled?: boolean;
-    /** RB-398-b ([2105] ANG-5): set (with `compacted:false`) when the pass PASSED the trigger gate
+    /** RB-398-b: set (with `compacted:false`) when the pass PASSED the trigger gate
      *  (natural or forced) but found nothing to summarize — no valid cut point / empty history. Was
      *  previously indistinguishable from the quiet under-threshold return, so a forced pass with
      *  nothing to do looked identical to one that never wanted to run; the caller's outcome event
@@ -502,7 +502,7 @@ export declare function maybeCompact(opts: MaybeCompactOptions): Promise<{
     firstKeptEntryId?: string;
     /** design/84 Seam C: true when this compaction reused a `summaryProvider` summary (no LLM call). */
     reused?: boolean;
-    /** Blackboard 2026-07-03 (ask ②): the working files re-read into the summary, in attachment order —
+    /** The working files re-read into the summary, in attachment order —
      *  surfaced so the `compacted` wire event carries them and a shell renders CC's post-compact
      *  `Read {path} (…)` cards. Set only when `compacted` is true and attachments were added. */
     attachedFiles?: Array<{
@@ -517,7 +517,7 @@ export declare function maybeCompact(opts: MaybeCompactOptions): Promise<{
      *  sanitized auto threshold — the "couldn't compact it down" signal {@link nextTrimForceBackoff}
      *  consumes. Set only when `compacted` and the post measurement succeeded. */
     postOverThreshold?: boolean;
-    /** TB telemetry B4 (service [397]): wall time of the whole compaction pass (context build → summary
+    /** TB telemetry B4: wall time of the whole compaction pass (context build → summary
      *  call → history rewrite → post measurement). Set only when `compacted` — the `compacted` wire event
      *  carries it so a consumer sees what a boundary pause actually cost. */
     durationMs?: number;
@@ -536,7 +536,7 @@ export declare function maybeCompact(opts: MaybeCompactOptions): Promise<{
      *  batch that adds an additive passthrough next to runtask's existing
      *  `...(comp.durationMs !== undefined ? { durationMs: comp.durationMs } : {})` spread. */
     phaseDurations?: CompactionPhaseDurations;
-    /** service [398] C3: set (with `compacted:false`) when the pass was over the auto threshold but the
+    /** C3: set (with `compacted:false`) when the pass was over the auto threshold but the
      *  anti-thrash floor (`minTokens`) suppressed it — the caller bridges this to `compaction.suppressed`. */
     suppressedByFloor?: {
         estTokens: number;
@@ -569,7 +569,7 @@ export declare function maybeCompact(opts: MaybeCompactOptions): Promise<{
         modelId?: string;
     };
     /**
-     * RB-126 / blackboard [1669]⑥ (2026-07-25, server's ask): the trigger inputs THIS pass used, reported on
+     * RB-126 (2026-07-25): the trigger inputs THIS pass used, reported on
      * EVERY return path — including the ordinary "nothing to do" one. A consumer rendering "how full is the
      * context" was otherwise forced to re-derive the numerator itself, and a re-derivation that drifts from
      * the engine's own is the projection-antipattern this repo keeps paying for: the displayed percentage