@wix/pathgrade 1.0.24 → 1.0.26

package/dist/sdk/scenario-machine-v2.js

@@ -0,0 +1,37 @@
+import { compileScenario as compileInternalScenario } from '../internal/direct-mcp-v2/compiler.js';
+import { canonicalJson, parseStrictJson } from '../internal/direct-mcp-v2/json.js';
+const V2_CONCURRENCY_POLICY = 'poison-effectful-overlap';
+function object(value) {
+    return value !== null && typeof value === 'object' && !Array.isArray(value);
+}
+/** Compile the public V2 source contract, whose concurrency safety policy is fixed by the version. */
+export function compileScenario(source) {
+    try {
+        canonicalJson(source);
+    }
+    catch (error) {
+        return { ok: false, diagnostics: [{ code: 'INVALID_JSON_VALUE', path: '$', message: error instanceof Error ? error.message : String(error) }] };
+    }
+    if (!object(source))
+        return compileInternalScenario(source);
+    const suppliedConcurrency = Object.prototype.hasOwnProperty.call(source, 'concurrency');
+    const compiled = compileInternalScenario({ ...source, concurrency: V2_CONCURRENCY_POLICY });
+    if (!suppliedConcurrency)
+        return compiled;
+    const diagnostic = { code: 'UNKNOWN_FIELD', path: '$/concurrency', message: 'unknown field concurrency' };
+    if (compiled.ok)
+        return { ok: false, diagnostics: [diagnostic] };
+    return {
+        ok: false,
+        diagnostics: [...compiled.diagnostics, diagnostic].sort((left, right) => canonicalJson(left).localeCompare(canonicalJson(right), 'en')),
+    };
+}
+/** Compile strict JSON text using the public V2 source contract. */
+export function compileScenarioText(text) {
+    try {
+        return compileScenario(parseStrictJson(text));
+    }
+    catch (error) {
+        return { ok: false, diagnostics: [{ code: 'INVALID_JSON_TEXT', path: '$', message: error instanceof Error ? error.message : String(error) }] };
+    }
+}

package/dist/sdk/snapshots.d.ts

@@ -1,6 +1,7 @@
 import type { LogEntry } from '../types.js';
 import type { ToolEvent } from '../tool-events.js';
 import type { AgentName, ConversationResult, Message, TurnTiming } from './types.js';
+import type { EvidenceEnvelope } from '../internal/direct-mcp-v2/types.js';
 export declare const RUN_SNAPSHOT_VERSION = 2;
 export interface RunSnapshot {
     version: 1 | 2;
@@ -17,6 +18,7 @@ export interface RunSnapshot {
         turnTimings: TurnTiming[];
     };
     workspace: string | null;
+    scenarioEvidence?: readonly EvidenceEnvelope[];
 }
 export declare function buildRunSnapshot(params: {
     agent: AgentName;
@@ -26,6 +28,7 @@ export declare function buildRunSnapshot(params: {
     workspace: string | null;
     timestamp?: string;
     sensitiveValues?: readonly string[];
+    scenarioEvidence?: readonly EvidenceEnvelope[];
 }): RunSnapshot;
 export declare class SnapshotParseError extends Error {
     constructor(message: string, options?: {

package/dist/sdk/snapshots.js

@@ -20,6 +20,7 @@ export function buildRunSnapshot(params) {
             turnTimings: [...conversationResult.turnTimings],
         },
         workspace,
+        ...(params.scenarioEvidence ? { scenarioEvidence: structuredClone(params.scenarioEvidence) } : {}),
     }, params.sensitiveValues);
 }
 export class SnapshotParseError extends Error {
@@ -153,5 +154,6 @@ function validateRunSnapshot(input) {
             turnTimings: snapshot.conversationResult.turnTimings,
         },
         workspace: typeof snapshot.workspace === 'string' ? snapshot.workspace : null,
+        scenarioEvidence: Array.isArray(snapshot.scenarioEvidence) ? snapshot.scenarioEvidence : [],
     };
 }

package/dist/sdk/types.d.ts

@@ -6,10 +6,12 @@ import type { DiagnosticsReport } from './diagnostics.js';
 import type { LLMPort } from '../utils/llm-types.js';
 import type { McpSafetyOptions } from './mcp-safety.js';
 import type { McpMockApprovalRule } from './mcp-mock-approvals.js';
+import type { EvidenceEnvelope } from '../internal/direct-mcp-v2/types.js';
+import type { ScenarioMachineV2 } from './scenario-machine-v2.js';
 export type AgentName = 'claude' | 'codex' | 'cursor' | 'opencode';
 export type AgentInteractionMode = 'prompt' | 'start_chat' | 'conversation';
 /** Runtime channel that actually executed the agent. */
-export type AgentExecutionTransport = AgentTransport | 'claude-agent-sdk' | 'cursor-agent';
+export type AgentExecutionTransport = AgentTransport | 'claude-agent-sdk' | 'cursor-agent' | 'acp';
 /** Privacy-safe execution dimensions exposed to result observers. */
 export interface AgentExecutionMetadata {
     name: AgentName;
@@ -29,6 +31,8 @@ export interface AgentOptions {
     copyFromHome?: string[];
     env?: Record<string, string>;
     mcpMock?: MockMcpServerDescriptor | MockMcpServerDescriptor[];
+    /** Host-authoritative direct MCP scenario for explicit Claude, Codex app-server, OpenCode, and admitted Cursor runtimes. */
+    mcpScenario?: ScenarioMachineV2;
     /** Ordered, persistent simulated-user decisions for approval-required generated MCP tools. */
     mcpMockApprovalRules?: McpMockApprovalRule[];
     mcpConfigFile?: string;
@@ -85,6 +89,8 @@ export interface Agent {
     readonly workspace: string;
     /** Provider/runtime metadata only; never contains prompts or paths. */
     readonly executionMetadata?: AgentExecutionMetadata;
+    /** Immutable canonical ScenarioMachineV2 evidence. Pathgrade agents return an empty array when no scenario is configured. */
+    readonly scenarioEvidence?: readonly EvidenceEnvelope[];
     dispose(): Promise<void>;
 }
 export interface ConverseOptions {
@@ -337,6 +343,8 @@ export interface ScorerContext {
     toolEvents: import('../tool-events.js').ToolEvent[];
     runCommand: (cmd: string) => Promise<CommandResult>;
     artifacts: SessionArtifacts;
+    /** Canonical host evidence for ScenarioMachineV2 scorers. */
+    scenarioEvidence?: readonly EvidenceEnvelope[];
 }
 export interface EvaluateOptions {
     failFast?: boolean;

package/dist/types.d.ts

@@ -175,6 +175,8 @@ export interface TrialResult {
      */
     total_cost_usd?: number;
     session_log: LogEntry[];
+    /** Redacted canonical ScenarioMachineV2 evidence for offline scoring and audit. */
+    scenario_evidence?: readonly import('./internal/direct-mcp-v2/types.js').EvidenceEnvelope[];
     skills_used?: string[];
     diagnostics?: DiagnosticsReport;
     result_kind?: import('./sdk/types.js').EvaluationResultKind;
@@ -390,6 +392,10 @@ export interface AgentSessionOptions {
     opencodeMcpToolNames?: string[];
     /** Managed-session-owned runtime for scripted generated MCP. */
     scriptedMcpHost?: import('./providers/scripted-mcp-mock-host.js').ScriptedMcpMockHost;
+    /** Parent-owned public ScenarioMachineV2 bridge; never serialized into child config. */
+    scenarioMcpServers?: Record<string, import('@anthropic-ai/claude-agent-sdk').McpSdkServerConfigWithInstance>;
+    /** Parent-owned admitted Codex app-server mount; projected only into thread/start memory. */
+    codexScenarioMount?: import('./agents/codex-app-server/scenario-mount.js').CodexScenarioMount;
     /** Runtime-only values inherited by drivers and persistence sinks for redaction. */
     sensitiveValues?: readonly string[];
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@wix/pathgrade",
-  "version": "1.0.24",
+  "version": "1.0.26",
   "packageManager": "yarn@4.12.0",
   "description": "Evaluate whether AI agents discover and use your skills correctly",
   "exports": {
@@ -140,5 +140,5 @@
     "typescript": "^5.9.3",
     "zod": "4.3.6"
   },
-  "falconPackageHash": "45feda0f936d0f7a5e2c85b4b55cd80ebe1f411b76a3c5c38b3fc149"
+  "falconPackageHash": "23390180e5694ad3ec546fb637b601045310b46b4d7f10927a63e0db"
 }