agent-afk 5.268.1 → 5.269.0

package/dist/whatif/probe-signflip.d.ts

@@ -0,0 +1,17 @@
+export declare const ZERO_TOLERANCE = 1e-9;
+export interface ProbeSignFlipResult {
+    nPaired: number;
+    nUnpaired: number;
+    nNonzero: number;
+    meanDelta: number;
+    probeDiffs: number[];
+    p: number | null;
+    minAchievableP: number | null;
+    underpoweredForSig: boolean;
+    method: 'exact' | 'montecarlo';
+}
+export interface ArmSamples {
+    baseline: number[];
+    candidate: number[];
+}
+export declare function computeProbeSignFlip(pairedEpisodeData: Map<string, ArmSamples>, episodeOrder: string[]): ProbeSignFlipResult;

package/dist/whatif/report.predictions.d.ts

@@ -0,0 +1,3 @@
+import type { Prediction, VerifiedPrediction } from './types.js';
+export declare function scoredOn(vp: VerifiedPrediction): string;
+export declare function renderVerifiedPredictionTable(predictions: Prediction[], verified: VerifiedPrediction[]): string[];

package/dist/whatif/report.signflip.d.ts

@@ -0,0 +1,4 @@
+import type { ArmImbalance } from './run.failures.js';
+import type { VerifiedPrediction } from './types.js';
+export declare function fmtP(p: number): string;
+export declare function renderProbeSignFlipSection(verified: VerifiedPrediction[], armImbalance?: ArmImbalance): string[];

package/dist/whatif/types.d.ts

@@ -196,6 +196,7 @@ export interface VerifiedPrediction {
     crossCheckAgreement?: number;
     crossCheckTooFew?: boolean;
     verdictReason?: 'judges disagree';
+    probeSignFlip?: import('./probe-signflip.js').ProbeSignFlipResult;
 }
 export interface DiscoveredDifference {
     id: string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-afk",
-  "version": "5.268.1",
+  "version": "5.269.0",
   "description": "Open-source coding-agent harness you can actually change — own the loop (prompts, gates, routing, skills, terminal states), use any model, run long tasks while you're away.",
   "main": "dist/index.mjs",
   "types": "dist/index.d.ts",