@nathapp/nax 0.44.0 → 0.45.0

package/dist/nax.js

@@ -3294,6 +3294,55 @@ var init_claude_complete = __esm(() => {
   };
 });
 
+// src/config/test-strategy.ts
+function resolveTestStrategy(raw) {
+  if (!raw)
+    return "test-after";
+  if (VALID_TEST_STRATEGIES.includes(raw))
+    return raw;
+  if (raw === "tdd")
+    return "tdd-simple";
+  if (raw === "three-session")
+    return "three-session-tdd";
+  if (raw === "tdd-lite")
+    return "three-session-tdd-lite";
+  return "test-after";
+}
+var VALID_TEST_STRATEGIES, COMPLEXITY_GUIDE = `## Complexity Classification Guide
+
+- simple: \u226450 LOC, single-file change, purely additive, no new dependencies \u2192 test-after
+- medium: 50\u2013200 LOC, 2\u20135 files, standard patterns, clear requirements \u2192 tdd-simple
+- complex: 200\u2013500 LOC, multiple modules, new abstractions or integrations \u2192 three-session-tdd
+- expert: 500+ LOC, architectural changes, cross-cutting concerns, high risk \u2192 three-session-tdd-lite
+
+### Security Override
+
+Security-critical functions (authentication, cryptography, tokens, sessions, credentials,
+password hashing, access control) must be classified at MINIMUM "medium" complexity
+regardless of LOC count. These require at minimum "tdd-simple" test strategy.`, TEST_STRATEGY_GUIDE = `## Test Strategy Guide
+
+- test-after: Simple changes with well-understood behavior. Write tests after implementation.
+- tdd-simple: Medium complexity. Write key tests first, implement, then fill coverage.
+- three-session-tdd: Complex stories. Full TDD cycle with separate test-writer and implementer sessions.
+- three-session-tdd-lite: Expert/high-risk stories. Full TDD with additional verifier session.`, GROUPING_RULES = `## Grouping Rules
+
+- Combine small, related tasks into a single "simple" or "medium" story.
+- Do NOT create separate stories for every single file or function unless complex.
+- Do NOT create standalone stories purely for test coverage or testing.
+  Each story's testStrategy already handles testing (tdd-simple writes tests first,
+  three-session-tdd uses separate test-writer session, test-after writes tests after).
+  Only create a dedicated test story for unique integration/E2E test logic that spans
+  multiple stories and cannot be covered by individual story test strategies.
+- Aim for coherent units of value. Maximum recommended stories: 10-15 per feature.`;
+var init_test_strategy = __esm(() => {
+  VALID_TEST_STRATEGIES = [
+    "test-after",
+    "tdd-simple",
+    "three-session-tdd",
+    "three-session-tdd-lite"
+  ];
+});
+
 // src/agents/claude-decompose.ts
 function buildDecomposePrompt(options) {
   return `You are a requirements analyst. Break down the following feature specification into user stories and classify each story's complexity.
@@ -3316,24 +3365,13 @@ Decompose this spec into user stories. For each story, provide:
 9. reasoning: Why this complexity level
 10. estimatedLOC: Estimated lines of code to change
 11. risks: Array of implementation risks
-12. testStrategy: "three-session-tdd" | "test-after"
+12. testStrategy: "test-after" | "tdd-simple" | "three-session-tdd" | "three-session-tdd-lite"
 
-testStrategy rules:
-- "three-session-tdd": ONLY for complex/expert tasks that are security-critical (auth, encryption, tokens, credentials) or define public API contracts consumers depend on
-- "test-after": for all other tasks including simple/medium complexity
-- A "simple" complexity task should almost never be "three-session-tdd"
+${COMPLEXITY_GUIDE}
 
-Complexity classification rules:
-- simple: 1-3 files, <100 LOC, straightforward implementation, existing patterns
-- medium: 3-6 files, 100-300 LOC, moderate logic, some new patterns
-- complex: 6+ files, 300-800 LOC, architectural changes, cross-cutting concerns
-- expert: Security/crypto/real-time/distributed systems, >800 LOC, new infrastructure
+${TEST_STRATEGY_GUIDE}
 
-Grouping Guidelines:
-- Combine small, related tasks (e.g., multiple utility functions, interfaces) into a single "simple" or "medium" story.
-- Do NOT create separate stories for every single file or function unless complex.
-- Aim for coherent units of value (e.g., "Implement User Authentication" vs "Create User Interface", "Create Login Service").
-- Maximum recommended stories: 10-15 per feature. Group aggressively if list grows too long.
+${GROUPING_RULES}
 
 Consider:
 1. Does infrastructure exist? (e.g., "add caching" when no cache layer exists = complex)
@@ -3402,7 +3440,7 @@ ${output.slice(0, 500)}`);
       reasoning: String(record.reasoning || "No reasoning provided"),
       estimatedLOC: Number(record.estimatedLOC) || 0,
       risks: Array.isArray(record.risks) ? record.risks : [],
-      testStrategy: record.testStrategy === "three-session-tdd" ? "three-session-tdd" : record.testStrategy === "test-after" ? "test-after" : undefined
+      testStrategy: resolveTestStrategy(typeof record.testStrategy === "string" ? record.testStrategy : undefined)
     };
   });
   if (stories.length === 0) {
@@ -3416,6 +3454,9 @@ function coerceComplexity(value) {
   }
   return "medium";
 }
+var init_claude_decompose = __esm(() => {
+  init_test_strategy();
+});
 
 // src/agents/cost.ts
 function parseTokenUsage(output) {
@@ -18398,6 +18439,7 @@ var init_claude = __esm(() => {
   init_pid_registry();
   init_logger2();
   init_claude_complete();
+  init_claude_decompose();
   init_claude_execution();
   init_claude_interactive();
   init_claude_plan();
@@ -19730,6 +19772,7 @@ class AcpAgentAdapter {
 var MAX_AGENT_OUTPUT_CHARS2 = 5000, MAX_RATE_LIMIT_RETRIES = 3, INTERACTION_TIMEOUT_MS, AGENT_REGISTRY, DEFAULT_ENTRY, _acpAdapterDeps, MAX_SESSION_AGE_MS;
 var init_adapter = __esm(() => {
   init_logger2();
+  init_claude_decompose();
   init_spawn_client();
   init_types2();
   init_cost2();
@@ -22042,7 +22085,7 @@ var package_default;
 var init_package = __esm(() => {
   package_default = {
     name: "@nathapp/nax",
-    version: "0.44.0",
+    version: "0.45.0",
     description: "AI Coding Agent Orchestrator \u2014 loops until done",
     type: "module",
     bin: {
@@ -22115,8 +22158,8 @@ var init_version = __esm(() => {
   NAX_VERSION = package_default.version;
   NAX_COMMIT = (() => {
     try {
-      if (/^[0-9a-f]{6,10}$/.test("05b2442"))
-        return "05b2442";
+      if (/^[0-9a-f]{6,10}$/.test("d6bdccb"))
+        return "d6bdccb";
     } catch {}
     try {
       const result = Bun.spawnSync(["git", "rev-parse", "--short", "HEAD"], {
@@ -25394,6 +25437,11 @@ async function buildStoryContextFull(prd, story, config2) {
 }
 function getAllReadyStories(prd) {
   const completedIds = new Set(prd.userStories.filter((s) => s.passes || s.status === "skipped").map((s) => s.id));
+  const logger = getSafeLogger2();
+  logger?.debug("routing", "getAllReadyStories: completed set", {
+    completedIds: [...completedIds],
+    totalStories: prd.userStories.length
+  });
   return prd.userStories.filter((s) => !s.passes && s.status !== "skipped" && s.status !== "failed" && s.status !== "paused" && s.status !== "blocked" && s.dependencies.every((dep) => completedIds.has(dep)));
 }
 var CONTEXT_MAX_TOKENS = 1e5, CONTEXT_RESERVED_TOKENS = 1e4;
@@ -31237,7 +31285,8 @@ async function executeFixStory(ctx, story, prd, iterations) {
     featureDir: ctx.featureDir,
     hooks: ctx.hooks,
     plugins: ctx.pluginRegistry,
-    storyStartTime: new Date().toISOString()
+    storyStartTime: new Date().toISOString(),
+    agentGetFn: ctx.agentGetFn
   };
   const result = await runPipeline(defaultPipeline, fixContext, ctx.eventEmitter);
   logger?.info("acceptance", `Fix story ${story.id} ${result.success ? "passed" : "failed"}`);
@@ -31273,7 +31322,8 @@ async function runAcceptanceLoop(ctx) {
       workdir: ctx.workdir,
       featureDir: ctx.featureDir,
       hooks: ctx.hooks,
-      plugins: ctx.pluginRegistry
+      plugins: ctx.pluginRegistry,
+      agentGetFn: ctx.agentGetFn
     };
     const { acceptanceStage: acceptanceStage2 } = await Promise.resolve().then(() => (init_acceptance2(), exports_acceptance));
     const acceptanceResult = await acceptanceStage2.execute(acceptanceContext);
@@ -32254,7 +32304,7 @@ function resolveMaxConcurrency(parallel) {
   }
   return Math.max(1, parallel);
 }
-async function executeParallel(stories, prdPath, projectRoot, config2, hooks, plugins, prd, featureDir, parallel, eventEmitter) {
+async function executeParallel(stories, prdPath, projectRoot, config2, hooks, plugins, prd, featureDir, parallel, eventEmitter, agentGetFn) {
   const logger = getSafeLogger();
   const maxConcurrency = resolveMaxConcurrency(parallel);
   const worktreeManager = new WorktreeManager;
@@ -32284,7 +32334,8 @@ async function executeParallel(stories, prdPath, projectRoot, config2, hooks, pl
       featureDir,
       hooks,
       plugins,
-      storyStartTime: new Date().toISOString()
+      storyStartTime: new Date().toISOString(),
+      agentGetFn
     };
     const worktreePaths = new Map;
     for (const story of batch) {
@@ -32658,7 +32709,7 @@ async function runParallelExecution(options, initialPrd) {
   const batchStoryMetrics = [];
   let conflictedStories = [];
   try {
-    const parallelResult = await _parallelExecutorDeps.executeParallel(readyStories, prdPath, workdir, config2, hooks, pluginRegistry, prd, featureDir, parallelCount, eventEmitter);
+    const parallelResult = await _parallelExecutorDeps.executeParallel(readyStories, prdPath, workdir, config2, hooks, pluginRegistry, prd, featureDir, parallelCount, eventEmitter, options.agentGetFn);
     const batchDurationMs = Date.now() - batchStartMs;
     const batchCompletedAt = new Date().toISOString();
     prd = parallelResult.updatedPrd;
@@ -65854,17 +65905,13 @@ init_registry();
 import { existsSync as existsSync9 } from "fs";
 import { join as join10 } from "path";
 import { createInterface } from "readline";
+init_test_strategy();
 init_pid_registry();
 init_logger2();
 
 // src/prd/schema.ts
+init_test_strategy();
 var VALID_COMPLEXITY = ["simple", "medium", "complex", "expert"];
-var VALID_TEST_STRATEGIES = [
-  "test-after",
-  "tdd-simple",
-  "three-session-tdd",
-  "three-session-tdd-lite"
-];
 var STORY_ID_NO_SEPARATOR = /^([A-Za-z]+)(\d+)$/;
 function extractJsonFromMarkdown(text) {
   const match = text.match(/```(?:json)?\s*\n([\s\S]*?)\n?\s*```/);
@@ -65934,9 +65981,7 @@ function validateStory(raw, index, allIds) {
     throw new Error(`[schema] story[${index}].routing.complexity "${rawComplexity}" is invalid. Valid values: ${VALID_COMPLEXITY.join(", ")}`);
   }
   const rawTestStrategy = routing.testStrategy ?? s.testStrategy;
-  const STRATEGY_ALIASES = { "tdd-lite": "three-session-tdd-lite" };
-  const normalizedStrategy = typeof rawTestStrategy === "string" ? STRATEGY_ALIASES[rawTestStrategy] ?? rawTestStrategy : rawTestStrategy;
-  const testStrategy = normalizedStrategy !== undefined && VALID_TEST_STRATEGIES.includes(normalizedStrategy) ? normalizedStrategy : "tdd-simple";
+  const testStrategy = resolveTestStrategy(typeof rawTestStrategy === "string" ? rawTestStrategy : undefined);
   const rawDeps = s.dependencies;
   const dependencies = Array.isArray(rawDeps) ? rawDeps : [];
   for (const dep of dependencies) {
@@ -66203,19 +66248,11 @@ Generate a JSON object with this exact structure (no markdown, no explanation \u
   ]
 }
 
-## Complexity Classification Guide
+${COMPLEXITY_GUIDE}
 
-- simple: \u226450 LOC, single-file change, purely additive, no new dependencies \u2192 test-after
-- medium: 50\u2013200 LOC, 2\u20135 files, standard patterns, clear requirements \u2192 tdd-simple
-- complex: 200\u2013500 LOC, multiple modules, new abstractions or integrations \u2192 three-session-tdd
-- expert: 500+ LOC, architectural changes, cross-cutting concerns, high risk \u2192 three-session-tdd-lite
+${TEST_STRATEGY_GUIDE}
 
-## Test Strategy Guide
-
-- test-after: Simple changes with well-understood behavior. Write tests after implementation.
-- tdd-simple: Medium complexity. Write key tests first, implement, then fill coverage.
-- three-session-tdd: Complex stories. Full TDD cycle with separate test-writer and implementer sessions.
-- three-session-tdd-lite: Expert/high-risk stories. Full TDD with additional verifier session.
+${GROUPING_RULES}
 
 ${outputFilePath ? `Write the PRD JSON directly to this file path: ${outputFilePath}
 Do NOT output the JSON to the conversation. Write the file, then reply with a brief confirmation.` : "Output ONLY the JSON object. Do not wrap in markdown code blocks."}`;
@@ -69227,9 +69264,20 @@ async function runExecutionPhase(options, prd, pluginRegistry) {
     batchingEnabled: options.useBatch
   });
   clearCache();
-  const batchPlan = options.useBatch ? precomputeBatchPlan(getAllReadyStories(prd), 4) : [];
+  const readyStories = getAllReadyStories(prd);
+  logger?.debug("routing", "Ready stories for batch routing", {
+    readyCount: readyStories.length,
+    readyIds: readyStories.map((s) => s.id),
+    allStories: prd.userStories.map((s) => ({
+      id: s.id,
+      status: s.status,
+      passes: s.passes,
+      deps: s.dependencies
+    }))
+  });
+  const batchPlan = options.useBatch ? precomputeBatchPlan(readyStories, 4) : [];
   if (options.useBatch) {
-    await tryLlmBatchRoute(options.config, getAllReadyStories(prd), "routing");
+    await tryLlmBatchRoute(options.config, readyStories, "routing");
   }
   if (options.parallel !== undefined) {
     const runParallelExecution2 = options.runParallelExecution ?? (await Promise.resolve().then(() => (init_parallel_executor(), exports_parallel_executor))).runParallelExecution;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nathapp/nax",
-  "version": "0.44.0",
+  "version": "0.45.0",
   "description": "AI Coding Agent Orchestrator — loops until done",
   "type": "module",
   "bin": {

package/src/agents/claude-decompose.ts

@@ -5,6 +5,7 @@
  * parseDecomposeOutput(), validateComplexity()
  */
 
+import { COMPLEXITY_GUIDE, GROUPING_RULES, TEST_STRATEGY_GUIDE, resolveTestStrategy } from "../config/test-strategy";
 import type { DecomposeOptions, DecomposeResult, DecomposedStory } from "./types";
 
 /**
@@ -31,24 +32,13 @@ Decompose this spec into user stories. For each story, provide:
 9. reasoning: Why this complexity level
 10. estimatedLOC: Estimated lines of code to change
 11. risks: Array of implementation risks
-12. testStrategy: "three-session-tdd" | "test-after"
+12. testStrategy: "test-after" | "tdd-simple" | "three-session-tdd" | "three-session-tdd-lite"
 
-testStrategy rules:
-- "three-session-tdd": ONLY for complex/expert tasks that are security-critical (auth, encryption, tokens, credentials) or define public API contracts consumers depend on
-- "test-after": for all other tasks including simple/medium complexity
-- A "simple" complexity task should almost never be "three-session-tdd"
+${COMPLEXITY_GUIDE}
 
-Complexity classification rules:
-- simple: 1-3 files, <100 LOC, straightforward implementation, existing patterns
-- medium: 3-6 files, 100-300 LOC, moderate logic, some new patterns
-- complex: 6+ files, 300-800 LOC, architectural changes, cross-cutting concerns
-- expert: Security/crypto/real-time/distributed systems, >800 LOC, new infrastructure
+${TEST_STRATEGY_GUIDE}
 
-Grouping Guidelines:
-- Combine small, related tasks (e.g., multiple utility functions, interfaces) into a single "simple" or "medium" story.
-- Do NOT create separate stories for every single file or function unless complex.
-- Aim for coherent units of value (e.g., "Implement User Authentication" vs "Create User Interface", "Create Login Service").
-- Maximum recommended stories: 10-15 per feature. Group aggressively if list grows too long.
+${GROUPING_RULES}
 
 Consider:
 1. Does infrastructure exist? (e.g., "add caching" when no cache layer exists = complex)
@@ -141,12 +131,7 @@ export function parseDecomposeOutput(output: string): DecomposedStory[] {
       reasoning: String(record.reasoning || "No reasoning provided"),
       estimatedLOC: Number(record.estimatedLOC) || 0,
       risks: Array.isArray(record.risks) ? record.risks : [],
-      testStrategy:
-        record.testStrategy === "three-session-tdd"
-          ? "three-session-tdd"
-          : record.testStrategy === "test-after"
-            ? "test-after"
-            : undefined,
+      testStrategy: resolveTestStrategy(typeof record.testStrategy === "string" ? record.testStrategy : undefined),
     };
   });
 

package/src/agents/types-extended.ts

@@ -117,7 +117,7 @@ export interface DecomposedStory {
   /** Implementation risks */
   risks: string[];
   /** Test strategy recommendation from LLM */
-  testStrategy?: "three-session-tdd" | "test-after";
+  testStrategy?: import("../config/test-strategy").TestStrategy;
 }
 
 /**

package/src/cli/plan.ts

@@ -16,6 +16,7 @@ import { scanCodebase } from "../analyze/scanner";
 import type { CodebaseScan } from "../analyze/types";
 import type { NaxConfig } from "../config";
 import { resolvePermissions } from "../config/permissions";
+import { COMPLEXITY_GUIDE, GROUPING_RULES, TEST_STRATEGY_GUIDE } from "../config/test-strategy";
 import { PidRegistry } from "../execution/pid-registry";
 import { getLogger } from "../logger";
 import { validatePlanOutput } from "../prd/schema";
@@ -320,19 +321,11 @@ Generate a JSON object with this exact structure (no markdown, no explanation
   ]
 }
 
-## Complexity Classification Guide
+${COMPLEXITY_GUIDE}
 
-- simple: ≤50 LOC, single-file change, purely additive, no new dependencies → test-after
-- medium: 50–200 LOC, 2–5 files, standard patterns, clear requirements → tdd-simple
-- complex: 200–500 LOC, multiple modules, new abstractions or integrations → three-session-tdd
-- expert: 500+ LOC, architectural changes, cross-cutting concerns, high risk → three-session-tdd-lite
+${TEST_STRATEGY_GUIDE}
 
-## Test Strategy Guide
-
-- test-after: Simple changes with well-understood behavior. Write tests after implementation.
-- tdd-simple: Medium complexity. Write key tests first, implement, then fill coverage.
-- three-session-tdd: Complex stories. Full TDD cycle with separate test-writer and implementer sessions.
-- three-session-tdd-lite: Expert/high-risk stories. Full TDD with additional verifier session.
+${GROUPING_RULES}
 
 ${
   outputFilePath

package/src/config/test-strategy.ts

@@ -0,0 +1,70 @@
+/**
+ * Test Strategy — Single Source of Truth
+ *
+ * Defines all valid test strategies, the normalizer, and shared prompt
+ * fragments used by plan.ts and claude-decompose.ts.
+ */
+
+import type { TestStrategy } from "./schema-types";
+
+// ─── Re-export type ───────────────────────────────────────────────────────────
+
+export type { TestStrategy };
+
+// ─── Valid values ─────────────────────────────────────────────────────────────
+
+export const VALID_TEST_STRATEGIES: readonly TestStrategy[] = [
+  "test-after",
+  "tdd-simple",
+  "three-session-tdd",
+  "three-session-tdd-lite",
+];
+
+// ─── Resolver ────────────────────────────────────────────────────────────────
+
+/**
+ * Validate and normalize a test strategy string.
+ * Returns a valid TestStrategy or falls back to "test-after".
+ */
+export function resolveTestStrategy(raw: string | undefined): TestStrategy {
+  if (!raw) return "test-after";
+  if (VALID_TEST_STRATEGIES.includes(raw as TestStrategy)) return raw as TestStrategy;
+  // Map legacy/typo values
+  if (raw === "tdd") return "tdd-simple";
+  if (raw === "three-session") return "three-session-tdd";
+  if (raw === "tdd-lite") return "three-session-tdd-lite";
+  return "test-after"; // safe fallback
+}
+
+// ─── Prompt fragments (shared by plan.ts and claude-decompose.ts) ────────────
+
+export const COMPLEXITY_GUIDE = `## Complexity Classification Guide
+
+- simple: ≤50 LOC, single-file change, purely additive, no new dependencies → test-after
+- medium: 50–200 LOC, 2–5 files, standard patterns, clear requirements → tdd-simple
+- complex: 200–500 LOC, multiple modules, new abstractions or integrations → three-session-tdd
+- expert: 500+ LOC, architectural changes, cross-cutting concerns, high risk → three-session-tdd-lite
+
+### Security Override
+
+Security-critical functions (authentication, cryptography, tokens, sessions, credentials,
+password hashing, access control) must be classified at MINIMUM "medium" complexity
+regardless of LOC count. These require at minimum "tdd-simple" test strategy.`;
+
+export const TEST_STRATEGY_GUIDE = `## Test Strategy Guide
+
+- test-after: Simple changes with well-understood behavior. Write tests after implementation.
+- tdd-simple: Medium complexity. Write key tests first, implement, then fill coverage.
+- three-session-tdd: Complex stories. Full TDD cycle with separate test-writer and implementer sessions.
+- three-session-tdd-lite: Expert/high-risk stories. Full TDD with additional verifier session.`;
+
+export const GROUPING_RULES = `## Grouping Rules
+
+- Combine small, related tasks into a single "simple" or "medium" story.
+- Do NOT create separate stories for every single file or function unless complex.
+- Do NOT create standalone stories purely for test coverage or testing.
+  Each story's testStrategy already handles testing (tdd-simple writes tests first,
+  three-session-tdd uses separate test-writer session, test-after writes tests after).
+  Only create a dedicated test story for unique integration/E2E test logic that spans
+  multiple stories and cannot be covered by individual story test strategies.
+- Aim for coherent units of value. Maximum recommended stories: 10-15 per feature.`;

package/src/execution/lifecycle/acceptance-loop.ts

@@ -143,6 +143,7 @@ async function executeFixStory(
     hooks: ctx.hooks,
     plugins: ctx.pluginRegistry,
     storyStartTime: new Date().toISOString(),
+    agentGetFn: ctx.agentGetFn,
   };
   const result = await runPipeline(defaultPipeline, fixContext, ctx.eventEmitter);
   logger?.info("acceptance", `Fix story ${story.id} ${result.success ? "passed" : "failed"}`);
@@ -189,6 +190,7 @@ export async function runAcceptanceLoop(ctx: AcceptanceLoopContext): Promise<Acc
       featureDir: ctx.featureDir,
       hooks: ctx.hooks,
       plugins: ctx.pluginRegistry,
+      agentGetFn: ctx.agentGetFn,
     };
 
     const { acceptanceStage } = await import("../../pipeline/stages/acceptance");

package/src/execution/parallel-coordinator.ts

@@ -8,7 +8,7 @@ import type { NaxConfig } from "../config";
 import type { LoadedHooksConfig } from "../hooks";
 import { getSafeLogger } from "../logger";
 import type { PipelineEventEmitter } from "../pipeline/events";
-import type { PipelineContext } from "../pipeline/types";
+import type { AgentGetFn } from "../pipeline/types";
 import type { PluginRegistry } from "../plugins/registry";
 import type { PRD, UserStory } from "../prd";
 import { markStoryFailed, markStoryPassed, savePRD } from "../prd";
@@ -108,6 +108,7 @@ export async function executeParallel(
   featureDir: string | undefined,
   parallel: number,
   eventEmitter?: PipelineEventEmitter,
+  agentGetFn?: AgentGetFn,
 ): Promise<{
   storiesCompleted: number;
   totalCost: number;
@@ -152,6 +153,7 @@ export async function executeParallel(
       hooks,
       plugins,
       storyStartTime: new Date().toISOString(),
+      agentGetFn,
     };
 
     // Create worktrees for all stories in batch

package/src/execution/parallel-executor.ts

@@ -17,6 +17,7 @@ import { fireHook } from "../hooks";
 import { getSafeLogger } from "../logger";
 import type { StoryMetrics } from "../metrics";
 import type { PipelineEventEmitter } from "../pipeline/events";
+import type { AgentGetFn } from "../pipeline/types";
 import type { PluginRegistry } from "../plugins/registry";
 import type { PRD } from "../prd";
 import { countStories, isComplete } from "../prd";
@@ -57,6 +58,7 @@ export interface ParallelExecutorOptions {
   pluginRegistry: PluginRegistry;
   formatterMode: "quiet" | "normal" | "verbose" | "json";
   headless: boolean;
+  agentGetFn?: AgentGetFn;
 }
 
 export interface RectificationStats {
@@ -158,6 +160,7 @@ export async function runParallelExecution(
       featureDir,
       parallelCount,
       eventEmitter,
+      options.agentGetFn,
     );
 
     const batchDurationMs = Date.now() - batchStartMs;

package/src/execution/runner-execution.ts

@@ -129,10 +129,24 @@ export async function runExecutionPhase(
   clearLlmCache();
 
   // PERF-1: Precompute batch plan once from ready stories
-  const batchPlan = options.useBatch ? precomputeBatchPlan(getAllReadyStories(prd), 4) : [];
+  const readyStories = getAllReadyStories(prd);
+
+  // BUG-068: debug log to diagnose unexpected storyCount in batch routing
+  logger?.debug("routing", "Ready stories for batch routing", {
+    readyCount: readyStories.length,
+    readyIds: readyStories.map((s) => s.id),
+    allStories: prd.userStories.map((s) => ({
+      id: s.id,
+      status: s.status,
+      passes: s.passes,
+      deps: s.dependencies,
+    })),
+  });
+
+  const batchPlan = options.useBatch ? precomputeBatchPlan(readyStories, 4) : [];
 
   if (options.useBatch) {
-    await tryLlmBatchRoute(options.config, getAllReadyStories(prd), "routing");
+    await tryLlmBatchRoute(options.config, readyStories, "routing");
   }
 
   // Parallel Execution Path (when --parallel is set)

package/src/execution/story-context.ts

@@ -175,6 +175,12 @@ export async function buildStoryContextFull(
 export function getAllReadyStories(prd: PRD): UserStory[] {
   const completedIds = new Set(prd.userStories.filter((s) => s.passes || s.status === "skipped").map((s) => s.id));
 
+  const logger = getSafeLogger();
+  logger?.debug("routing", "getAllReadyStories: completed set", {
+    completedIds: [...completedIds],
+    totalStories: prd.userStories.length,
+  });
+
   return prd.userStories.filter(
     (s) =>
       !s.passes &&

package/src/prd/schema.ts

@@ -5,6 +5,7 @@
  */
 
 import type { Complexity, TestStrategy } from "../config";
+import { resolveTestStrategy } from "../config/test-strategy";
 import type { PRD, UserStory } from "./types";
 import { validateStoryId } from "./validate";
 
@@ -13,12 +14,6 @@ import { validateStoryId } from "./validate";
 // ---------------------------------------------------------------------------
 
 const VALID_COMPLEXITY: Complexity[] = ["simple", "medium", "complex", "expert"];
-const VALID_TEST_STRATEGIES: TestStrategy[] = [
-  "test-after",
-  "tdd-simple",
-  "three-session-tdd",
-  "three-session-tdd-lite",
-];
 
 /** Pattern matching ST001 → ST-001 style IDs (prefix letters + digits, no separator) */
 const STORY_ID_NO_SEPARATOR = /^([A-Za-z]+)(\d+)$/;
@@ -140,15 +135,10 @@ function validateStory(raw: unknown, index: number, allIds: Set<string>): UserSt
   }
 
   // testStrategy — accept from routing.testStrategy or top-level testStrategy
-  // Also map legacy/LLM-hallucinated aliases: tdd-lite → tdd-simple
   const rawTestStrategy = routing.testStrategy ?? s.testStrategy;
-  const STRATEGY_ALIASES: Record<string, TestStrategy> = { "tdd-lite": "three-session-tdd-lite" };
-  const normalizedStrategy =
-    typeof rawTestStrategy === "string" ? (STRATEGY_ALIASES[rawTestStrategy] ?? rawTestStrategy) : rawTestStrategy;
-  const testStrategy: TestStrategy =
-    normalizedStrategy !== undefined && (VALID_TEST_STRATEGIES as unknown[]).includes(normalizedStrategy)
-      ? (normalizedStrategy as TestStrategy)
-      : "tdd-simple";
+  const testStrategy: TestStrategy = resolveTestStrategy(
+    typeof rawTestStrategy === "string" ? rawTestStrategy : undefined,
+  );
 
   // dependencies
   const rawDeps = s.dependencies;