aiwcli 0.15.7 → 0.17.0

package/dist/templates/cc-native/_cc-native/{plan-review/lib → lib-ts}/review-pipeline.ts

@@ -1,543 +1,511 @@
-/**
- * Review pipeline: orchestrates the full plan review lifecycle.
- * Wires together plan-discovery, settings, agent-selection, graduation,
- * output-builder, and existing modules (orchestrator, corroboration, etc.).
- */
-
-import * as fs from "node:fs";
-import * as path from "node:path";
-
-import { resolveMandatoryAgents, assignModelsToAgents, selectAgents } from "./agent-selection.js";
-import { computeCorroboratedDecision } from "./corroboration.js";
-import { computePassEligible, extractTopIssuesForTracker, advanceIterationState } from "./graduation.js";
-import { runOrchestrator } from "./orchestrator.js";
-import { truncateAgentIssues, overrideVerdictsByThreshold, buildReviewOutput } from "./output-builder.js";
-import { runPlanQuestions } from "./plan-questions.js";
-import { runPreflight } from "./preflight.js";
-import { runAgentReview } from "./reviewers/index.js";
-import { getContextBySessionId, getAllContexts } from "../../../_core/lib-ts/context/context-store.js";
-import { logDiagnostic } from "../../../_core/lib-ts/hooks/hook-utils.js";
-import { getContextReviewsDir, getContextDir, getReviewFolderPath } from "../../../_core/lib-ts/runtime/constants.js";
-import {
-  logDebug,
-  logInfo,
-  logWarn,
-  logError,
-} from "../../../_core/lib-ts/runtime/logger.js";
-import { eprint } from "../../../_core/lib-ts/runtime/utils.js";
-import type { ContextState } from "../../../_core/lib-ts/types.js";
-import { writeCombinedArtifacts, buildCorroborationReport, buildHighIssuesDocument, writeReviewTracker } from "../../artifacts/lib/index.js";
-import type { ReviewTrackerEntry } from "../../artifacts/lib/index.js";
-import {
-  isPlanAlreadyReviewed,
-  wasPlanPreviouslyDenied,
-  getLastPlanReview,
-  markPlanReviewed,
-  wasPlanQuestionsAgentAsked,
-  markQuestionsAsked,
-  resetPlanQuestionsAsked,
-} from "../../lib-ts/cc-native-state.js";
-import { debugLog } from "../../lib-ts/debug.js";
-import { discoverPlan } from "../../lib-ts/plan-discovery.js";
-import { loadSettings, loadModelsConfig, loadAgentLibrary, DEFAULT_ORCHESTRATOR } from "../../lib-ts/settings.js";
-import { DEFAULT_REVIEW_ITERATIONS, loadIterationState, saveIterationState } from "../../lib-ts/state.js";
-import type {
-  AgentConfig,
-  AgentReviewSettings,
-  LoadedSettings,
-  OrchestratorConfig,
-  PlanReviewSettings,
-  ReviewerResult,
-  CombinedReviewResult,
-  OrchestratorResult,
-  IterationState,
-  PipelineInput,
-  PipelineResult,
-} from "../../lib-ts/types.js";
-import { REVIEW_SCHEMA } from "../../lib-ts/types.js";
-
-const HOOK = "review-pipeline";
-
-// ---------------------------------------------------------------------------
-// Context Lookup (private — only used here)
-// ---------------------------------------------------------------------------
-
-function getActiveContextForReview(sessionId: string, projectRoot: string): ContextState | null {
-  const ctx = getContextBySessionId(sessionId, projectRoot);
-  if (ctx) {
-    logInfo(HOOK, `Found context by session_id: ${ctx.id}`);
-    return ctx;
-  }
-  const allActive = getAllContexts("active", projectRoot);
-  const planning = allActive.filter(c => c.mode === "active" || c.mode === "has_staged_work");
-  if (planning.length === 1) {
-    logInfo(HOOK, `Found single planning context: ${planning[0]!.id}`);
-    return planning[0]!;
-  }
-  if (planning.length > 1) {
-    logWarn(HOOK, `Multiple planning contexts (${planning.length}), cannot determine which to use`);
-  } else if (allActive.length > 0) {
-    logInfo(HOOK, `Found ${allActive.length} active context(s) but none in planning mode`);
-  } else {
-    logInfo(HOOK, "No active contexts found");
-  }
-  return null;
-}
-
-// ---------------------------------------------------------------------------
-// Pipeline
-// ---------------------------------------------------------------------------
-
-export async function runReviewPipeline(input: PipelineInput): Promise<PipelineResult> {
-  const { sessionId, base, aiwcliDir, transcriptPath, payload } = input;
-
-  // 1. Load settings
-  const settings: LoadedSettings = loadSettings(aiwcliDir);
-  const planSettings: PlanReviewSettings = settings.planReview;
-  const agentSettings: AgentReviewSettings = settings.agentReview;
-
-  const planReviewEnabled = planSettings.enabled ?? true;
-  const agentReviewEnabled = agentSettings.enabled ?? true;
-
-  if (!planReviewEnabled && !agentReviewEnabled) {
-    return { action: "skip", reason: "Both plan and agent review disabled" };
-  }
-
-  // 2. Discover plan
-  const discovered = discoverPlan(transcriptPath);
-  if (!discovered) {
-    return { action: "skip", reason: "No plan file found in ~/.claude/plans/. The plan may not have been written yet." };
-  }
-
-  const { content: plan, hash: planHash, path: planPath } = discovered;
-
-  // 3. Find active context (moved before questions gate for plan path change detection)
-  const activeContext = getActiveContextForReview(sessionId, base);
-  if (!activeContext) {
-    return { action: "skip", reason: "No active planning context found for this session." };
-  }
-
-  const contextId = activeContext.id;
-  const reviewsDir = path.join(getContextReviewsDir(contextId, base), "cc-native");
-  const contextPath = getContextDir(contextId, base);
-
-  // 4a. Load iteration state
-  let iterationState: IterationState | null = loadIterationState(reviewsDir);
-
-  // 4a-migration. Backfill sessionId for old iteration files
-  if (iterationState && !iterationState.sessionId) {
-    logInfo(HOOK, `Migrating iteration state: adding sessionId=${sessionId}`);
-    iterationState.sessionId = sessionId;
-    saveIterationState(reviewsDir, iterationState);  // Persist migration
-  }
-
-  // 4a-session. Detect session change — reset iteration state for new planning session
-  if (iterationState && iterationState.sessionId && iterationState.sessionId !== sessionId) {
-    logInfo(HOOK, `Session changed (${iterationState.sessionId} → ${sessionId}), resetting iteration state`);
-    iterationState = null;  // Force fresh state creation below
-  }
-
-  // 4a-init. Initialize if null
-  if (!iterationState) {
-    iterationState = {
-      current: 1, max: 1, complexity: "medium",
-      history: [], graduated: [], passStreaks: {},
-      lastPlanHash: "", lastPlanPath: "",
-      sessionId,
-    };
-    saveIterationState(reviewsDir, iterationState);
-  }
-
-  // 4b. Detect plan file change — reset iteration state for new plan topic
-  const lastPath = iterationState.lastPlanPath ?? "";
-  if (lastPath && lastPath !== planPath) {
-    logInfo(HOOK, `Plan file changed (${path.basename(lastPath)}→${path.basename(planPath)}), resetting iteration state for new plan`);
-    iterationState = {
-      current: 1, max: 1, complexity: "medium",
-      history: [], graduated: [], passStreaks: {},
-      lastPlanHash: "", lastPlanPath: "",
-      sessionId,
-    };
-    saveIterationState(reviewsDir, iterationState);
-    resetPlanQuestionsAsked(sessionId, base);
-  }
-
-  // 5. Questions gate
-  if (!wasPlanQuestionsAgentAsked(sessionId, base)) {
-    logInfo(HOOK, "Questions gate: plan-questions agent has not run yet, running now");
-    const questionsTimeout = agentSettings.timeout ?? 120;
-    const questionsResult = await runPlanQuestions(plan, aiwcliDir, questionsTimeout, undefined, sessionId);
-
-    markQuestionsAsked(sessionId, base, "agent");
-
-    const hasQuestions = questionsResult && (
-      questionsResult.questions.length > 0 ||
-      questionsResult.assumptions.length > 0 ||
-      questionsResult.ambiguities.length > 0
-    );
-
-    if (hasQuestions) {
-      const questionsList = questionsResult.questions.map((q: string, i: number) => `${i + 1}. ${q}`).join("\n");
-      const assumptionsList = questionsResult.assumptions.length > 0
-        ? `\n\nAssumptions detected:\n${questionsResult.assumptions.map((a: string) => `- ${a}`).join("\n")}`
-        : "";
-      const ambiguitiesList = questionsResult.ambiguities.length > 0
-        ? `\n\nAmbiguities detected:\n${questionsResult.ambiguities.map((a: string) => `- ${a}`).join("\n")}`
-        : "";
-      const contextMsg = `## Plan Questions (from independent review)\n\nAn agent reviewed your plan in a fresh context — without access to your session history or codebase exploration. It identified these questions:\n\n${questionsList}${assumptionsList}${ambiguitiesList}\n\nAsk the user these questions using AskUserQuestion before submitting the plan.`;
-      return {
-        action: "block",
-        contextText: contextMsg,
-        blockReason: "Ask the user clarifying questions before submitting the plan. Use AskUserQuestion with the questions above.",
-      };
-    }
-
-    logInfo(HOOK, "Questions gate: no questions generated, proceeding to review");
-  } else {
-    logInfo(HOOK, "Questions gate: agent already ran, skipping");
-  }
-
-  // 6. Hash + dedup
-  logDiagnostic(HOOK, "receive", `plan_size=${plan.length}, session=${sessionId.slice(0, 8)}`, {
-    inputs: { plan_hash: planHash, plan_size: plan.length, session_id: sessionId.slice(0, 12) },
-  });
-
-  // Plan-hash deduplication
-  logDebug(HOOK, `Plan hash: ${planHash}`);
-  if (isPlanAlreadyReviewed(sessionId, planHash, base)) {
-    const lastReview = getLastPlanReview(sessionId, planHash, base);
-
-    if (wasPlanPreviouslyDenied(sessionId, planHash, base)) {
-      return {
-        action: "block",
-        contextText: "[Plan Review] Plan content unchanged since last review which found issues.",
-        blockReason: "Plan unchanged since denial. Modify the plan to address review findings, then attempt ExitPlanMode again.",
-      };
-    } 
-      const verdict = lastReview?.iteration?.latest_verdict || "pass";
-      return { action: "skip", reason: `Plan already reviewed (verdict: ${verdict}). Skipping re-review.` };
-    
-  }
-
-  // 7. Iteration bounds check
-  if (iterationState.current > iterationState.max) {
-    return { action: "skip", reason: `Max review iterations reached (${iterationState.current - 1}/${iterationState.max}), allowing plan through.` };
-  }
-
-  // Initialize result containers
-  let orchResult: OrchestratorResult | null = null;
-  const agentResults: Record<string, ReviewerResult> = {};
-  let detectedComplexity = "medium";
-
-  // Preflight: validate provider+model combos before committing agents or orchestrator
-  const preflightEnabled = agentSettings.preflight?.enabled ?? true;
-  let preflightAvailable: Map<string, Set<string>> | undefined;
-
-  if (preflightEnabled && agentReviewEnabled) {
-    logInfo(HOOK, "=== PREFLIGHT: Checking provider availability ===");
-    const preflightTimeoutMs = agentSettings.preflight?.timeoutMs;
-    const modelsConfig = loadModelsConfig(settings);
-    const preflightReport = await runPreflight(modelsConfig, preflightTimeoutMs);
-
-    if (preflightReport.allFailed) {
-      logWarn(HOOK, "All providers failed preflight checks");
-      // Preflight failures skip review rather than block because an unavailable
-      // reviewer is worse than no reviewer. A permanently broken config will
-      // silently pass all plans — mitigated by the log warnings above.
-      eprint("[plan-review] All AI providers unavailable. Skipping review.");
-      return { action: "skip", reason: "No AI providers passed preflight. Check CLI, API keys, model access, and quota." };
-    }
-
-    preflightAvailable = preflightReport.available;
-  }
-
-  // 7. PHASE 1: Orchestrator
-  const graduatedSet = new Set(iterationState.graduated);
-  if (graduatedSet.size > 0) {
-    logInfo(HOOK, `Graduated agents from previous iterations: ${[...graduatedSet].sort().join(", ")}`);
-  }
-
-  const agentLibrary = agentReviewEnabled ? loadAgentLibrary(aiwcliDir, agentSettings) : [];
-  const originalAgentCount = agentLibrary.length;
-  const enabledAgents = agentLibrary.filter(a => !graduatedSet.has(a.name));
-  const timeout = agentSettings.timeout ?? 120;
-  const legacyMode = agentSettings.legacyMode === true;
-
-  const orchSettings = agentSettings.orchestrator ?? DEFAULT_ORCHESTRATOR;
-  const orchestratorConfig: OrchestratorConfig = {
-    enabled: (orchSettings.enabled ?? true) && agentReviewEnabled,
-    model: orchSettings.model ?? "haiku",
-    provider: orchSettings.provider,
-    timeout: orchSettings.timeout ?? 30,
-  };
-
-  const mandatoryConfig = agentSettings.mandatoryAgents ?? ["handoff-readiness", "clarity-auditor", "skeptic"];
-  const alwaysMandatory = resolveMandatoryAgents(mandatoryConfig, "simple");
-
-  logDebug(HOOK, `Agent library: ${agentLibrary.map(a => a.name)}`);
-  logDebug(HOOK, `Mandatory agents: ${[...alwaysMandatory].sort()}`);
-  logDebug(HOOK, `Orchestrator enabled: ${orchestratorConfig.enabled}`);
-
-  const phase1Promises: Array<{ name: string; promise: Promise<ReviewerResult | OrchestratorResult> }> = [];
-
-  if (orchestratorConfig.enabled && enabledAgents.length > 0 && !legacyMode) {
-    // Guard orchestrator against preflight failures (always uses claude provider)
-    const orchProvider = orchestratorConfig.provider ?? "claude";
-    const orchModel = orchestratorConfig.model;
-    const orchPassed = !preflightAvailable ||
-      (preflightAvailable.has(orchProvider) && preflightAvailable.get(orchProvider)!.has(orchModel));
-
-    if (!orchPassed) {
-      logWarn(HOOK, `Orchestrator model ${orchProvider}:${orchModel} failed preflight, skipping`);
-    } else {
-      phase1Promises.push({
-        name: "orchestrator",
-        promise: runOrchestrator(plan, enabledAgents, orchestratorConfig, agentSettings, alwaysMandatory),
-      });
-    }
-  }
-
-  logInfo(HOOK, `=== PHASE 1: Running ${phase1Promises.length} tasks in parallel ===`);
-
-  if (phase1Promises.length > 0) {
-    const results = await Promise.allSettled(
-      phase1Promises.map(async ({ name, promise }) => {
-        const result = await promise;
-        return { name, result };
-      }),
-    );
-    for (const [i, r] of results.entries()) {
-      if (r.status === "fulfilled") {
-        if (r.value.name === "orchestrator") orchResult = r.value.result as OrchestratorResult;
-        logInfo(HOOK, `${r.value.name} completed`);
-      } else {
-        const failedName = phase1Promises[i]?.name ?? "unknown";
-        logError(HOOK, `${failedName} failed: ${r.reason}`);
-      }
-    }
-  }
-
-  // 8. PHASE 2: Agent Selection
-  let selectedAgents: AgentConfig[] = [];
-  let mandatoryNames = alwaysMandatory;
-
-  if (agentReviewEnabled) {
-    logInfo(HOOK, "=== PHASE 2: Agent Selection ===");
-
-    const selectionResult = selectAgents({
-      enabledAgents,
-      orchResult,
-      mandatoryConfig,
-      agentSettings,
-      legacyMode,
-    });
-
-    selectedAgents = selectionResult.selectedAgents;
-    mandatoryNames = selectionResult.mandatoryNames;
-    detectedComplexity = selectionResult.detectedComplexity;
-
-    logDiagnostic(HOOK, "decide", `Selected ${selectedAgents.length} agents, complexity=${detectedComplexity}`, {
-      decision: "agents_selected",
-      reasoning: `orchestrator=${orchResult !== null}, legacy=${legacyMode}`,
-      inputs: {
-        agents: selectedAgents.map(a => a.name),
-        complexity: detectedComplexity,
-        mandatory_count: selectedAgents.filter(a => mandatoryNames.has(a.name)).length,
-      },
-    });
-
-    // Update iteration state with complexity/max
-    const reviewIterations: Record<string, number> = {
-      ...DEFAULT_REVIEW_ITERATIONS,
-      ...agentSettings.reviewIterations,
-    };
-    iterationState.complexity = detectedComplexity;
-    iterationState.max = reviewIterations[detectedComplexity] ?? iterationState.max;
-    logDebug(HOOK, `Iteration state: ${iterationState.current}/${iterationState.max} (${detectedComplexity})`);
-
-    // Assign providers + models (filtered by preflight results if available)
-    const modelsConfig = loadModelsConfig(settings);
-    selectedAgents = assignModelsToAgents(selectedAgents, modelsConfig, preflightAvailable);
-    logInfo(HOOK, `Model assignments: ${selectedAgents.map(a => `${a.name}→${a.provider}:${a.model}`).join(", ")}`);
-
-    // 9. PHASE 3: Run agents
-    if (selectedAgents.length > 0) {
-      logInfo(HOOK, "=== PHASE 3: Agent Reviews ===");
-      logInfo(HOOK, `Launching ${selectedAgents.length} agents in parallel`);
-
-      debugLog(contextPath, sessionId, "hook", "agent_review_start", {
-        agents: selectedAgents.map(a => a.name),
-        timeout,
-        complexity: detectedComplexity,
-      });
-
-      const agentPromises = selectedAgents.map(async agent => {
-        const result = await runAgentReview(plan, agent, REVIEW_SCHEMA, timeout, contextPath, sessionId);
-        return { agent, result };
-      });
-
-      const agentSettled = await Promise.allSettled(agentPromises);
-      for (const [i, r] of agentSettled.entries()) {
-        if (r.status === "fulfilled") {
-          const { agent, result } = r.value;
-          agentResults[agent.name] = result;
-          logInfo(HOOK, `${agent.name} completed with verdict: ${result.verdict}`);
-        } else {
-          const failedAgent = selectedAgents[i]!;
-          logError(HOOK, `${failedAgent.name} failed with exception: ${r.reason}`);
-          agentResults[failedAgent.name] = {
-            name: failedAgent.name,
-            ok: false,
-            verdict: "error",
-            data: {},
-            raw: "",
-            err: String(r.reason),
-          };
-        }
-      }
-    }
-  }
-
-  // 10. Issue truncation + verdict override
-  const maxIssuesPerAgent = agentSettings.maxIssuesPerAgent ?? 3;
-  truncateAgentIssues(agentResults, maxIssuesPerAgent);
-
-  const passEligible = computePassEligible(agentResults);
-  if (passEligible.length > 0) {
-    logInfo(HOOK, `Pass-eligible agents this iteration: ${passEligible.join(", ")}`);
-  }
-
-  const highIssueThreshold = agentSettings.highIssueThreshold ?? 3;
-  overrideVerdictsByThreshold(agentResults, highIssueThreshold);
-
-  // PHASE 4: Generate Output
-  logInfo(HOOK, "=== PHASE 4: Generate Output ===");
-
-  if (Object.keys(agentResults).length === 0) {
-    if (graduatedSet.size > 0 && originalAgentCount > 0) {
-      return { action: "skip", reason: "All agent reviewers graduated from previous iterations — no review needed." };
-    }
-    return { action: "skip", reason: "All reviewers failed to produce results. Check stderr logs for details." };
-  }
-
-  // 11. Corroboration
-  const corroborationResult = computeCorroboratedDecision(agentResults);
-  const overall = corroborationResult.verdict;
-
-  const combinedResult: CombinedReviewResult = {
-    plan_hash: planHash,
-    overall_verdict: overall,
-    orchestration: orchResult,
-    agents: agentResults,
-    timestamp: new Date().toISOString(),
-  };
-
-  const displaySettings = {
-    ...planSettings.display,
-    ...agentSettings.display,
-  };
-  const combinedSettings = { display: displaySettings };
-
-  const currentIteration = iterationState.current;
-
-  // 12. Write artifacts
-  const reviewFolder = getReviewFolderPath(contextId, currentIteration, base);
-  fs.mkdirSync(reviewFolder, { recursive: true });
-  logInfo(HOOK, `Created review folder: ${reviewFolder}`);
-
-  const reviewFile = writeCombinedArtifacts(
-    base, plan, combinedResult, payload, combinedSettings,
-    undefined, reviewFolder, currentIteration, corroborationResult,
-  );
-  logInfo(HOOK, `Saved review: ${reviewFile}`);
-
-  const corroborationReport = buildCorroborationReport(corroborationResult);
-  const corroborationPath = path.join(reviewFolder, "corroboration.md");
-  fs.writeFileSync(corroborationPath, corroborationReport, "utf-8");
-  logInfo(HOOK, `Saved corroboration report: ${corroborationPath}`);
-
-  try {
-    fs.writeFileSync(path.join(reviewFolder, "plan.md"), plan, "utf-8");
-    logDebug(HOOK, `Saved plan snapshot: ${path.join(reviewFolder, "plan.md")}`);
-  } catch (error) {
-    logWarn(HOOK, `Failed to save plan snapshot: ${error}`);
-  }
-
-  // Build high-issues document
-  const highIssuesDoc = buildHighIssuesDocument(combinedResult, corroborationResult);
-  const highIssuesPath = path.join(reviewFolder, "high-issues.md");
-  fs.writeFileSync(highIssuesPath, highIssuesDoc, "utf-8");
-
-  // 15. Build output
-  const shouldDeny = corroborationResult.blocking.length > 0;
-  const reviewScore = shouldDeny ? 1 : 0;
-  const denyReason = shouldDeny ? "corroborated_issues" : "no_corroboration";
-
-  logInfo(HOOK, `REVIEW_DECISION: verdict=${combinedResult.overall_verdict}, deny=${shouldDeny}, reason=${denyReason}, score=${reviewScore.toFixed(2)}`);
-  logDiagnostic(HOOK, "result", `verdict=${combinedResult.overall_verdict}, deny=${shouldDeny}, reason=${denyReason}`, {
-    decision: shouldDeny ? "deny" : "allow",
-    reasoning: `reason=${denyReason}, score=${reviewScore.toFixed(2)}`,
-    inputs: {
-      overall_verdict: combinedResult.overall_verdict,
-      review_score: Math.round(reviewScore * 100) / 100,
-      agent_count: Object.keys(agentResults).length,
-    },
-  });
-
-  // Terminal progress
-  const verdictEmoji = shouldDeny ? "❌" : "✅";
-  eprint(`[plan-review] ${verdictEmoji} ${combinedResult.overall_verdict.toUpperCase()} (score=${reviewScore.toFixed(2)})`);
-  if (shouldDeny) {
-    eprint(`[plan-review] Blocking ExitPlanMode — ${denyReason}`);
-  }
-
-  // 13. Advance iteration
-  const advancement = advanceIterationState(
-    iterationState, planHash, planPath, overall, shouldDeny, passEligible, agentResults,
-  );
-  iterationState = advancement.updatedState;
-  if (advancement.newGraduates.length > 0) {
-    logInfo(HOOK, `Newly graduated (2 consecutive passes): ${advancement.newGraduates.join(", ")}`);
-  }
-
-  // 14. Save iteration state
-  saveIterationState(reviewsDir, iterationState);
-
-  // Write review tracker
-  const ccNativeReviewsDir = path.dirname(reviewFolder);
-  const trackerDecision = shouldDeny ? "blocked" : "allow";
-  const topIssuesList = extractTopIssuesForTracker(combinedResult, 5);
-  const trackerEntry: ReviewTrackerEntry = {
-    iteration: currentIteration,
-    timestamp: new Date().toISOString().replace("T", " ").slice(0, 16),
-    planHash,
-    verdict: combinedResult.overall_verdict,
-    decision: trackerDecision,
-    score: reviewScore,
-    topIssues: topIssuesList,
-    reviewFolder,
-  };
-  writeReviewTracker(ccNativeReviewsDir, trackerEntry);
-  logInfo(HOOK, `Updated review tracker: ${path.join(ccNativeReviewsDir, "review-tracker.md")}`);
-
-  // Build final output
-  const output = buildReviewOutput({
-    combinedResult,
-    corroborationResult,
-    iterationState: { ...iterationState, current: currentIteration },
-    reviewFile,
-    highIssuesPath,
-  });
-
-  // Mark plan reviewed
-  const disposition = shouldDeny
-    ? `hook_deny_iter_${currentIteration}`
-    : `hook_allow_iter_${currentIteration}`;
-  markPlanReviewed(sessionId, planHash, base, HOOK, { ...iterationState, current: currentIteration }, disposition);
-
-  return {
-    action: "block",
-    contextText: output.contextText,
-    blockReason: output.blockReason,
-  };
-}
+/**
+ * Review pipeline: orchestrates the full plan review lifecycle.
+ * Wires together plan-discovery, settings, agent-selection, graduation,
+ * output-builder, and existing modules (orchestrator, corroboration, etc.).
+ */
+
+import * as fs from "node:fs";
+import * as path from "node:path";
+
+import {
+  logDebug,
+  logInfo,
+  logWarn,
+  logError,
+} from "../../_core/lib-ts/runtime/logger.js";
+import { logDiagnostic } from "../../_core/lib-ts/hooks/hook-utils.js";
+import { eprint } from "../../_core/lib-ts/runtime/utils.js";
+import { getContextReviewsDir, getContextDir, getReviewFolderPath } from "../../_core/lib-ts/runtime/constants.js";
+import { getContextBySessionId, getAllContexts } from "../../_core/lib-ts/context/context-store.js";
+
+import type {
+  AgentConfig,
+  OrchestratorConfig,
+  ReviewerResult,
+  CombinedReviewResult,
+  OrchestratorResult,
+  IterationState,
+  PipelineInput,
+  PipelineResult,
+} from "./types.js";
+import { REVIEW_SCHEMA } from "./types.js";
+import type { ContextState } from "../../_core/lib-ts/types.js";
+
+import { discoverPlan } from "./plan-discovery.js";
+import { loadSettings, loadModelsConfig, loadAgentLibrary, DEFAULT_ORCHESTRATOR } from "./settings.js";
+import { resolveMandatoryAgents, assignModelsToAgents, selectAgents } from "./agent-selection.js";
+import { computePassEligible, extractTopIssuesForTracker, advanceIterationState } from "./graduation.js";
+import { truncateAgentIssues, overrideVerdictsByThreshold, buildReviewOutput } from "./output-builder.js";
+import { DEFAULT_REVIEW_ITERATIONS, loadIterationState, saveIterationState } from "./state.js";
+
+import {
+  isPlanAlreadyReviewed,
+  wasPlanPreviouslyDenied,
+  getLastPlanReview,
+  markPlanReviewed,
+  wasPlanQuestionsAgentAsked,
+  markQuestionsAsked,
+  resetPlanQuestionsAsked,
+} from "./cc-native-state.js";
+
+import { computeCorroboratedDecision } from "./corroboration.js";
+import { runOrchestrator } from "./orchestrator.js";
+import { debugLog } from "./debug.js";
+import { writeCombinedArtifacts, buildCorroborationReport, buildHighIssuesDocument, writeReviewTracker } from "./artifacts.js";
+import type { ReviewTrackerEntry } from "./artifacts.js";
+import { runAgentReview } from "./reviewers/index.js";
+import { runPlanQuestions } from "./plan-questions.js";
+
+const HOOK = "review-pipeline";
+
+// ---------------------------------------------------------------------------
+// Context Lookup (private — only used here)
+// ---------------------------------------------------------------------------
+
+function getActiveContextForReview(sessionId: string, projectRoot: string): ContextState | null {
+  const ctx = getContextBySessionId(sessionId, projectRoot);
+  if (ctx) {
+    logInfo(HOOK, `Found context by session_id: ${ctx.id}`);
+    return ctx;
+  }
+  const allActive = getAllContexts("active", projectRoot);
+  const planning = allActive.filter(c => c.mode === "active" || c.mode === "has_plan");
+  if (planning.length === 1) {
+    logInfo(HOOK, `Found single planning context: ${planning[0]!.id}`);
+    return planning[0]!;
+  }
+  if (planning.length > 1) {
+    logWarn(HOOK, `Multiple planning contexts (${planning.length}), cannot determine which to use`);
+  } else if (allActive.length > 0) {
+    logInfo(HOOK, `Found ${allActive.length} active context(s) but none in planning mode`);
+  } else {
+    logInfo(HOOK, "No active contexts found");
+  }
+  return null;
+}
+
+// ---------------------------------------------------------------------------
+// Pipeline
+// ---------------------------------------------------------------------------
+
+export async function runReviewPipeline(input: PipelineInput): Promise<PipelineResult> {
+  const { sessionId, base, aiwcliDir, transcriptPath, payload } = input;
+
+  // 1. Load settings
+  const settings = loadSettings(aiwcliDir);
+  const planSettings = settings.planReview ?? {};
+  const agentSettings = settings.agentReview ?? {};
+
+  const planReviewEnabled = planSettings.enabled ?? true;
+  const agentReviewEnabled = agentSettings.enabled ?? true;
+
+  if (!planReviewEnabled && !agentReviewEnabled) {
+    return { action: "skip", reason: "Both plan and agent review disabled" };
+  }
+
+  // 2. Discover plan
+  const discovered = discoverPlan(transcriptPath);
+  if (!discovered) {
+    return { action: "skip", reason: "No plan file found in ~/.claude/plans/. The plan may not have been written yet." };
+  }
+
+  const { content: plan, hash: planHash, path: planPath } = discovered;
+
+  // 3. Find active context (moved before questions gate for plan path change detection)
+  const activeContext = getActiveContextForReview(sessionId, base);
+  if (!activeContext) {
+    return { action: "skip", reason: "No active planning context found for this session." };
+  }
+
+  const contextId = activeContext.id;
+  const reviewsDir = path.join(getContextReviewsDir(contextId, base), "cc-native");
+  const contextPath = getContextDir(contextId, base);
+
+  // 4a. Load iteration state
+  let iterationState: IterationState | null = loadIterationState(reviewsDir);
+
+  // 4a-migration. Backfill sessionId for old iteration files
+  if (iterationState && !iterationState.sessionId) {
+    logInfo(HOOK, `Migrating iteration state: adding sessionId=${sessionId}`);
+    iterationState.sessionId = sessionId;
+    saveIterationState(reviewsDir, iterationState);  // Persist migration
+  }
+
+  // 4a-session. Detect session change — reset iteration state for new planning session
+  if (iterationState && iterationState.sessionId && iterationState.sessionId !== sessionId) {
+    logInfo(HOOK, `Session changed (${iterationState.sessionId} → ${sessionId}), resetting iteration state`);
+    iterationState = null;  // Force fresh state creation below
+  }
+
+  // 4a-init. Initialize if null
+  if (!iterationState) {
+    iterationState = {
+      current: 1, max: 1, complexity: "medium",
+      history: [], graduated: [], passStreaks: {},
+      lastPlanHash: "", lastPlanPath: "",
+      sessionId: sessionId,
+    };
+    saveIterationState(reviewsDir, iterationState);
+  }
+
+  // 4b. Detect plan file change — reset iteration state for new plan topic
+  const lastPath = iterationState.lastPlanPath ?? "";
+  if (lastPath && lastPath !== planPath) {
+    logInfo(HOOK, `Plan file changed (${path.basename(lastPath)}→${path.basename(planPath)}), resetting iteration state for new plan`);
+    iterationState = {
+      current: 1, max: 1, complexity: "medium",
+      history: [], graduated: [], passStreaks: {},
+      lastPlanHash: "", lastPlanPath: "",
+      sessionId: sessionId,
+    };
+    saveIterationState(reviewsDir, iterationState);
+    resetPlanQuestionsAsked(sessionId, base);
+  }
+
+  // 5. Questions gate
+  if (!wasPlanQuestionsAgentAsked(sessionId, base)) {
+    logInfo(HOOK, "Questions gate: plan-questions agent has not run yet, running now");
+    const timeout = typeof agentSettings.timeout === "number" ? agentSettings.timeout : 120;
+    const questionsResult = await runPlanQuestions(plan, aiwcliDir, timeout, undefined, sessionId);
+
+    markQuestionsAsked(sessionId, base, "agent");
+
+    const hasQuestions = questionsResult && (
+      questionsResult.questions.length > 0 ||
+      questionsResult.assumptions.length > 0 ||
+      questionsResult.ambiguities.length > 0
+    );
+
+    if (hasQuestions) {
+      const questionsList = questionsResult.questions.map((q: string, i: number) => `${i + 1}. ${q}`).join("\n");
+      const assumptionsList = questionsResult.assumptions.length > 0
+        ? `\n\nAssumptions detected:\n${questionsResult.assumptions.map((a: string) => `- ${a}`).join("\n")}`
+        : "";
+      const ambiguitiesList = questionsResult.ambiguities.length > 0
+        ? `\n\nAmbiguities detected:\n${questionsResult.ambiguities.map((a: string) => `- ${a}`).join("\n")}`
+        : "";
+      const contextMsg = `## Plan Questions (from independent review)\n\nAn agent reviewed your plan in a fresh context — without access to your session history or codebase exploration. It identified these questions:\n\n${questionsList}${assumptionsList}${ambiguitiesList}\n\nAsk the user these questions using AskUserQuestion before submitting the plan.`;
+      return {
+        action: "block",
+        contextText: contextMsg,
+        blockReason: "Ask the user clarifying questions before submitting the plan. Use AskUserQuestion with the questions above.",
+      };
+    }
+
+    logInfo(HOOK, "Questions gate: no questions generated, proceeding to review");
+  } else {
+    logInfo(HOOK, "Questions gate: agent already ran, skipping");
+  }
+
+  // 6. Hash + dedup
+  logDiagnostic(HOOK, "receive", `plan_size=${plan.length}, session=${sessionId.slice(0, 8)}`, {
+    inputs: { plan_hash: planHash, plan_size: plan.length, session_id: sessionId.slice(0, 12) },
+  });
+
+  // Plan-hash deduplication
+  logDebug(HOOK, `Plan hash: ${planHash}`);
+  if (isPlanAlreadyReviewed(sessionId, planHash, base)) {
+    const lastReview = getLastPlanReview(sessionId, planHash, base);
+
+    if (wasPlanPreviouslyDenied(sessionId, planHash, base)) {
+      return {
+        action: "block",
+        contextText: "[Plan Review] Plan content unchanged since last review which found issues.",
+        blockReason: "Plan unchanged since denial. Modify the plan to address review findings, then attempt ExitPlanMode again.",
+      };
+    } else {
+      const verdict = lastReview?.iteration?.latest_verdict || "pass";
+      return { action: "skip", reason: `Plan already reviewed (verdict: ${verdict}). Skipping re-review.` };
+    }
+  }
+
+  // 7. Iteration bounds check
+  if (iterationState.current > iterationState.max) {
+    return { action: "skip", reason: `Max review iterations reached (${iterationState.current - 1}/${iterationState.max}), allowing plan through.` };
+  }
+
+  // Initialize result containers
+  let orchResult: OrchestratorResult | null = null;
+  const agentResults: Record<string, ReviewerResult> = {};
+  let detectedComplexity = "medium";
+
+  // 7. PHASE 1: Orchestrator
+  const graduatedSet = new Set(iterationState.graduated);
+  if (graduatedSet.size > 0) {
+    logInfo(HOOK, `Graduated agents from previous iterations: ${[...graduatedSet].sort().join(", ")}`);
+  }
+
+  const agentLibrary = agentReviewEnabled ? loadAgentLibrary(aiwcliDir, agentSettings) : [];
+  const originalAgentCount = agentLibrary.length;
+  const enabledAgents = agentLibrary.filter(a => !graduatedSet.has(a.name));
+  const timeout = typeof agentSettings.timeout === "number" ? agentSettings.timeout : 120;
+  const legacyMode = agentSettings.legacyMode === true;
+
+  const orchSettings = agentSettings.orchestrator ?? DEFAULT_ORCHESTRATOR;
+  const orchestratorConfig: OrchestratorConfig = {
+    enabled: (orchSettings.enabled ?? true) && agentReviewEnabled,
+    model: orchSettings.model ?? "haiku",
+    timeout: orchSettings.timeout ?? 30,
+  };
+
+  const mandatoryConfig = agentSettings.mandatoryAgents ?? ["handoff-readiness", "clarity-auditor", "skeptic"];
+  const alwaysMandatory = resolveMandatoryAgents(mandatoryConfig, "simple");
+
+  logDebug(HOOK, `Agent library: ${agentLibrary.map(a => a.name)}`);
+  logDebug(HOOK, `Mandatory agents: ${[...alwaysMandatory].sort()}`);
+  logDebug(HOOK, `Orchestrator enabled: ${orchestratorConfig.enabled}`);
+
+  const phase1Promises: Array<{ name: string; promise: Promise<ReviewerResult | OrchestratorResult> }> = [];
+
+  if (orchestratorConfig.enabled && enabledAgents.length > 0 && !legacyMode) {
+    phase1Promises.push({
+      name: "orchestrator",
+      promise: runOrchestrator(plan, enabledAgents, orchestratorConfig, agentSettings, alwaysMandatory),
+    });
+  }
+
+  logInfo(HOOK, `=== PHASE 1: Running ${phase1Promises.length} tasks in parallel ===`);
+
+  if (phase1Promises.length > 0) {
+    const results = await Promise.allSettled(
+      phase1Promises.map(async ({ name, promise }) => {
+        const result = await promise;
+        return { name, result };
+      }),
+    );
+    for (const [i, r] of results.entries()) {
+      if (r.status === "fulfilled") {
+        if (r.value.name === "orchestrator") orchResult = r.value.result as OrchestratorResult;
+        logInfo(HOOK, `${r.value.name} completed`);
+      } else {
+        const failedName = phase1Promises[i]?.name ?? "unknown";
+        logError(HOOK, `${failedName} failed: ${r.reason}`);
+      }
+    }
+  }
+
+  // 8. PHASE 2: Agent Selection
+  let selectedAgents: AgentConfig[] = [];
+  let mandatoryNames = alwaysMandatory;
+
+  if (agentReviewEnabled) {
+    logInfo(HOOK, "=== PHASE 2: Agent Selection ===");
+
+    const selectionResult = selectAgents({
+      enabledAgents,
+      orchResult,
+      mandatoryConfig,
+      agentSettings,
+      legacyMode,
+    });
+
+    selectedAgents = selectionResult.selectedAgents;
+    mandatoryNames = selectionResult.mandatoryNames;
+    detectedComplexity = selectionResult.detectedComplexity;
+
+    logDiagnostic(HOOK, "decide", `Selected ${selectedAgents.length} agents, complexity=${detectedComplexity}`, {
+      decision: "agents_selected",
+      reasoning: `orchestrator=${orchResult !== null}, legacy=${legacyMode}`,
+      inputs: {
+        agents: selectedAgents.map(a => a.name),
+        complexity: detectedComplexity,
+        mandatory_count: selectedAgents.filter(a => mandatoryNames.has(a.name)).length,
+      },
+    });
+
+    // Update iteration state with complexity/max
+    const reviewIterations: Record<string, number> = {
+      ...DEFAULT_REVIEW_ITERATIONS,
+      ...(agentSettings.reviewIterations ?? {}),
+    };
+    iterationState.complexity = detectedComplexity;
+    iterationState.max = reviewIterations[detectedComplexity] ?? iterationState.max;
+    logDebug(HOOK, `Iteration state: ${iterationState.current}/${iterationState.max} (${detectedComplexity})`);
+
+    // Assign random providers + models
+    const modelsConfig = loadModelsConfig(settings);
+    selectedAgents = assignModelsToAgents(selectedAgents, modelsConfig);
+    logInfo(HOOK, `Model assignments: ${selectedAgents.map(a => `${a.name}→${a.provider}:${a.model}`).join(", ")}`);
+
+    // 9. PHASE 3: Run agents
+    if (selectedAgents.length > 0) {
+      logInfo(HOOK, "=== PHASE 3: Agent Reviews ===");
+      logInfo(HOOK, `Launching ${selectedAgents.length} agents in parallel`);
+
+      debugLog(contextPath, sessionId, "hook", "agent_review_start", {
+        agents: selectedAgents.map(a => a.name),
+        timeout,
+        complexity: detectedComplexity,
+      });
+
+      const agentPromises = selectedAgents.map(async agent => {
+        const result = await runAgentReview(plan, agent, REVIEW_SCHEMA, timeout, contextPath, sessionId);
+        return { agent, result };
+      });
+
+      const agentSettled = await Promise.allSettled(agentPromises);
+      for (const [i, r] of agentSettled.entries()) {
+        if (r.status === "fulfilled") {
+          const { agent, result } = r.value;
+          agentResults[agent.name] = result;
+          logInfo(HOOK, `${agent.name} completed with verdict: ${result.verdict}`);
+        } else {
+          const failedAgent = selectedAgents[i]!;
+          logError(HOOK, `${failedAgent.name} failed with exception: ${r.reason}`);
+          agentResults[failedAgent.name] = {
+            name: failedAgent.name,
+            ok: false,
+            verdict: "error",
+            data: {},
+            raw: "",
+            err: String(r.reason),
+          };
+        }
+      }
+    }
+  }
+
+  // 10. Issue truncation + verdict override
+  const maxIssuesPerAgent = typeof agentSettings.maxIssuesPerAgent === "number"
+    ? agentSettings.maxIssuesPerAgent : 3;
+  truncateAgentIssues(agentResults, maxIssuesPerAgent);
+
+  const passEligible = computePassEligible(agentResults);
+  if (passEligible.length > 0) {
+    logInfo(HOOK, `Pass-eligible agents this iteration: ${passEligible.join(", ")}`);
+  }
+
+  const highIssueThreshold = typeof agentSettings.highIssueThreshold === "number" ? agentSettings.highIssueThreshold : 3;
+  overrideVerdictsByThreshold(agentResults, highIssueThreshold);
+
+  // PHASE 4: Generate Output
+  logInfo(HOOK, "=== PHASE 4: Generate Output ===");
+
+  if (Object.keys(agentResults).length === 0) {
+    if (graduatedSet.size > 0 && originalAgentCount > 0) {
+      return { action: "skip", reason: "All agent reviewers graduated from previous iterations — no review needed." };
+    }
+    return { action: "skip", reason: "All reviewers failed to produce results. Check stderr logs for details." };
+  }
+
+  // 11. Corroboration
+  const corroborationResult = computeCorroboratedDecision(agentResults);
+  const overall = corroborationResult.verdict;
+
+  const combinedResult: CombinedReviewResult = {
+    plan_hash: planHash,
+    overall_verdict: overall,
+    orchestration: orchResult,
+    agents: agentResults,
+    timestamp: new Date().toISOString(),
+  };
+
+  const displaySettings = {
+    ...(planSettings.display ?? {}),
+    ...(agentSettings.display ?? {}),
+  };
+  const combinedSettings = { display: displaySettings };
+
+  const currentIteration = iterationState.current;
+
+  // 12. Write artifacts
+  const reviewFolder = getReviewFolderPath(contextId, currentIteration, base);
+  fs.mkdirSync(reviewFolder, { recursive: true });
+  logInfo(HOOK, `Created review folder: ${reviewFolder}`);
+
+  const reviewFile = writeCombinedArtifacts(
+    base, plan, combinedResult, payload, combinedSettings,
+    undefined, reviewFolder, currentIteration, corroborationResult,
+  );
+  logInfo(HOOK, `Saved review: ${reviewFile}`);
+
+  const corroborationReport = buildCorroborationReport(corroborationResult);
+  const corroborationPath = path.join(reviewFolder, "corroboration.md");
+  fs.writeFileSync(corroborationPath, corroborationReport, "utf-8");
+  logInfo(HOOK, `Saved corroboration report: ${corroborationPath}`);
+
+  try {
+    fs.writeFileSync(path.join(reviewFolder, "plan.md"), plan, "utf-8");
+    logDebug(HOOK, `Saved plan snapshot: ${path.join(reviewFolder, "plan.md")}`);
+  } catch (e) {
+    logWarn(HOOK, `Failed to save plan snapshot: ${e}`);
+  }
+
+  // Build high-issues document
+  const highIssuesDoc = buildHighIssuesDocument(combinedResult, corroborationResult);
+  const highIssuesPath = path.join(reviewFolder, "high-issues.md");
+  fs.writeFileSync(highIssuesPath, highIssuesDoc, "utf-8");
+
+  // 15. Build output
+  const shouldDeny = corroborationResult.blocking.length > 0;
+  const reviewScore = shouldDeny ? 1.0 : 0.0;
+  const denyReason = shouldDeny ? "corroborated_issues" : "no_corroboration";
+
+  logInfo(HOOK, `REVIEW_DECISION: verdict=${combinedResult.overall_verdict}, deny=${shouldDeny}, reason=${denyReason}, score=${reviewScore.toFixed(2)}`);
+  logDiagnostic(HOOK, "result", `verdict=${combinedResult.overall_verdict}, deny=${shouldDeny}, reason=${denyReason}`, {
+    decision: shouldDeny ? "deny" : "allow",
+    reasoning: `reason=${denyReason}, score=${reviewScore.toFixed(2)}`,
+    inputs: {
+      overall_verdict: combinedResult.overall_verdict,
+      review_score: Math.round(reviewScore * 100) / 100,
+      agent_count: Object.keys(agentResults).length,
+    },
+  });
+
+  // Terminal progress
+  const verdictEmoji = shouldDeny ? "❌" : "✅";
+  eprint(`[plan-review] ${verdictEmoji} ${combinedResult.overall_verdict.toUpperCase()} (score=${reviewScore.toFixed(2)})`);
+  if (shouldDeny) {
+    eprint(`[plan-review] Blocking ExitPlanMode — ${denyReason}`);
+  }
+
+  // 13. Advance iteration
+  const advancement = advanceIterationState(
+    iterationState, planHash, planPath, overall, shouldDeny, passEligible, agentResults,
+  );
+  iterationState = advancement.updatedState;
+  if (advancement.newGraduates.length > 0) {
+    logInfo(HOOK, `Newly graduated (2 consecutive passes): ${advancement.newGraduates.join(", ")}`);
+  }
+
+  // 14. Save iteration state
+  saveIterationState(reviewsDir, iterationState);
+
+  // Write review tracker
+  const ccNativeReviewsDir = path.dirname(reviewFolder);
+  const trackerDecision = shouldDeny ? "blocked" : "allow";
+  const topIssuesList = extractTopIssuesForTracker(combinedResult, 5);
+  const trackerEntry: ReviewTrackerEntry = {
+    iteration: currentIteration,
+    timestamp: new Date().toISOString().replace("T", " ").slice(0, 16),
+    planHash,
+    verdict: combinedResult.overall_verdict,
+    decision: trackerDecision,
+    score: reviewScore,
+    topIssues: topIssuesList,
+    reviewFolder,
+  };
+  writeReviewTracker(ccNativeReviewsDir, trackerEntry);
+  logInfo(HOOK, `Updated review tracker: ${path.join(ccNativeReviewsDir, "review-tracker.md")}`);
+
+  // Build final output
+  const output = buildReviewOutput({
+    combinedResult,
+    corroborationResult,
+    iterationState: { ...iterationState, current: currentIteration },
+    reviewFile,
+    highIssuesPath,
+  });
+
+  // Mark plan reviewed
+  const disposition = shouldDeny
+    ? `hook_deny_iter_${currentIteration}`
+    : `hook_allow_iter_${currentIteration}`;
+  markPlanReviewed(sessionId, planHash, base, HOOK, { ...iterationState, current: currentIteration }, disposition);
+
+  return {
+    action: "block",
+    contextText: output.contextText,
+    blockReason: output.blockReason,
+  };
+}