role-os 2.0.0 → 2.1.0

package/src/dispatch.mjs

@@ -1,310 +1,333 @@
-/**
- * Runtime dispatch engine.
- *
- * Turns a staffed, validated chain into an executable dispatch manifest
- * that multi-claude can consume. Role-OS owns staffing/routing/evidence;
- * multi-claude owns execution.
- *
- * This module:
- * 1. Maps roles → role configs (tool profiles, system prompts, budgets)
- * 2. Tracks execution state (queued → running → completed/failed/blocked)
- * 3. Collects outputs back into the packet system
- * 4. Generates escalation packets for blocked/rejected work
- * 5. Closes chains to final completion
- */
-
-import { existsSync, mkdirSync, writeFileSync, readFileSync } from "node:fs";
-import { join, resolve } from "node:path";
-import { resolveBlocked, resolveRejected } from "./escalation.mjs";
-
-// ── Tool profiles per role ────────────────────────────────────────────────────
-// What tools each role is allowed to use. Sandboxing by contract.
-
-const TOOL_PROFILES = {
-  // Core
-  "Orchestrator":         ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
-  "Product Strategist":   ["Read", "Glob", "Grep", "Write"],
-  "Critic Reviewer":      ["Read", "Glob", "Grep", "Bash"],
-
-  // Design
-  "UI Designer":          ["Read", "Glob", "Grep", "Write", "Edit"],
-  "Brand Guardian":       ["Read", "Glob", "Grep"],
-
-  // Engineering
-  "Backend Engineer":     ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
-  "Frontend Developer":   ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
-  "Test Engineer":        ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
-  "Performance Engineer": ["Read", "Glob", "Grep", "Bash"],
-  "Refactor Engineer":    ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
-  "Security Reviewer":    ["Read", "Glob", "Grep", "Bash"],
-  "Dependency Auditor":   ["Read", "Glob", "Grep", "Bash"],
-
-  // Treatment
-  "Repo Researcher":      ["Read", "Glob", "Grep", "Bash"],
-  "Repo Translator":      ["Read", "Glob", "Grep", "Write", "Edit"],
-  "Docs Architect":       ["Read", "Glob", "Grep", "Write", "Edit"],
-  "Metadata Curator":     ["Read", "Glob", "Grep", "Write", "Edit"],
-  "Coverage Auditor":     ["Read", "Glob", "Grep", "Bash"],
-  "Deployment Verifier":  ["Read", "Glob", "Grep", "Bash"],
-  "Release Engineer":     ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
-
-  // Growth / Marketing
-  "Launch Strategist":    ["Read", "Glob", "Grep", "Write"],
-  "Content Strategist":   ["Read", "Glob", "Grep", "Write"],
-  "Community Manager":    ["Read", "Glob", "Grep", "Write"],
-  "Support Triage Lead":  ["Read", "Glob", "Grep", "Write"],
-  "Launch Copywriter":    ["Read", "Glob", "Grep", "Write", "Edit"],
-
-  // Product
-  "Feedback Synthesizer": ["Read", "Glob", "Grep"],
-  "Roadmap Prioritizer":  ["Read", "Glob", "Grep", "Write"],
-  "Spec Writer":          ["Read", "Glob", "Grep", "Write", "Edit"],
-
-  // Research
-  "UX Researcher":        ["Read", "Glob", "Grep"],
-  "Competitive Analyst":  ["Read", "Glob", "Grep"],
-  "Trend Researcher":     ["Read", "Glob", "Grep"],
-  "User Interview Synthesizer": ["Read", "Glob", "Grep"],
-};
-
-// ── Default role config ─────────────────────────────────────────────────────
-
-const DEFAULTS = {
-  model: "sonnet",
-  maxTurns: 30,
-  maxBudgetUsd: 5.0,
-  timeoutMs: 10 * 60 * 1000, // 10 minutes
-};
-
-// ── Execution states ──────────────────────────────────────────────────────────
-
-export const EXEC_STATES = [
-  "queued",         // in chain, not yet launched
-  "running",        // role session active
-  "waiting",        // waiting for upstream handoff
-  "blocked",        // hit a block condition
-  "needs-review",   // completed work, awaiting verdict
-  "completed",      // approved and done
-  "failed",         // session error or timeout
-  "interrupted",    // manually stopped
-];
-
-// ── System prompt builder ─────────────────────────────────────────────────────
-
-function buildRolePrompt(roleName, packetContent, chainContext) {
-  return `You are operating as ${roleName} in a Role-OS managed chain.
-
-## Your Role Contract
-Follow the ${roleName} contract exactly. Your quality bar, inputs, outputs, and escalation rules are defined in .claude/agents/.
-
-## Current Packet
-${packetContent}
-
-## Chain Context
-You are step ${chainContext.stepNumber} of ${chainContext.totalSteps} in this chain.
-${chainContext.previousRole ? `Previous role: ${chainContext.previousRole} (${chainContext.previousStatus})` : "You are the first role in this chain."}
-${chainContext.nextRole ? `Next role: ${chainContext.nextRole}` : "You are the last role before Critic review."}
-
-## Handoff Requirements
-When you finish, produce a structured handoff:
-1. Summary of what you did
-2. Artifacts produced (files, changes, docs)
-3. Open questions or risks for the next role
-4. Evidence items for your verdict (kind, reference, claim, status)
-
-## Stop Conditions
-- If you cannot proceed: output a BLOCKED status with a clear reason
-- If the upstream handoff is insufficient: output a BLOCKED status explaining what's missing
-- Do not silently skip work — surface every gap explicitly
-`;
-}
-
-// ── Dispatch manifest builder ─────────────────────────────────────────────────
-
-/**
- * Build a dispatch manifest from a routed chain.
- * This is the contract between Role-OS (planning) and multi-claude (execution).
- *
- * @param {Object} options
- * @param {string} options.packetFile - Path to the packet markdown
- * @param {string} options.packetContent - Packet markdown content
- * @param {Array<{role: {name: string, pack: string}}>} options.chainRoles - Assembled chain
- * @param {string} options.cwd - Working directory for execution
- * @param {Object} [options.overrides] - Per-role config overrides
- * @returns {DispatchManifest}
- */
-export function buildDispatchManifest({ packetFile, packetContent, chainRoles, cwd, overrides = {} }) {
-  const runId = `run-${Date.now()}`;
-  const steps = [];
-
-  for (let i = 0; i < chainRoles.length; i++) {
-    const { role } = chainRoles[i];
-    const roleOverrides = overrides[role.name] || {};
-
-    const chainContext = {
-      stepNumber: i + 1,
-      totalSteps: chainRoles.length,
-      previousRole: i > 0 ? chainRoles[i - 1].role.name : null,
-      previousStatus: i > 0 ? "pending" : null,
-      nextRole: i < chainRoles.length - 1 ? chainRoles[i + 1].role.name : null,
-    };
-
-    steps.push({
-      stepIndex: i,
-      packetId: `${runId}-step-${i}`,
-      role: role.name,
-      pack: role.pack,
-      tools: TOOL_PROFILES[role.name] || ["Read", "Glob", "Grep"],
-      systemPrompt: buildRolePrompt(role.name, packetContent, chainContext),
-      model: roleOverrides.model || DEFAULTS.model,
-      maxTurns: roleOverrides.maxTurns || DEFAULTS.maxTurns,
-      maxBudgetUsd: roleOverrides.maxBudgetUsd || DEFAULTS.maxBudgetUsd,
-      timeoutMs: roleOverrides.timeoutMs || DEFAULTS.timeoutMs,
-      state: "queued",
-      dependsOn: i > 0 ? [`${runId}-step-${i - 1}`] : [],
-      handoff: {
-        from: i > 0 ? chainRoles[i - 1].role.name : null,
-        to: i < chainRoles.length - 1 ? chainRoles[i + 1].role.name : null,
-      },
-    });
-  }
-
-  return {
-    version: 1,
-    runId,
-    createdAt: new Date().toISOString(),
-    packetFile: resolve(packetFile),
-    cwd: resolve(cwd),
-    totalSteps: steps.length,
-    steps,
-  };
-}
-
-// ── Execution state tracker ───────────────────────────────────────────────────
-
-/**
- * Update a step's execution state in the manifest.
- *
- * @param {DispatchManifest} manifest
- * @param {number} stepIndex
- * @param {string} newState - One of EXEC_STATES
- * @param {Object} [result] - Execution result data
- * @returns {DispatchManifest} Updated manifest (mutated in place)
- */
-export function updateStepState(manifest, stepIndex, newState, result = {}) {
-  const step = manifest.steps[stepIndex];
-  if (!step) throw new Error(`Invalid step index: ${stepIndex}`);
-  if (!EXEC_STATES.includes(newState)) throw new Error(`Invalid state: ${newState}`);
-
-  step.state = newState;
-  step.result = {
-    ...step.result,
-    ...result,
-    updatedAt: new Date().toISOString(),
-  };
-
-  // Auto-advance: if this step completed, mark next step as ready
-  if (newState === "completed" && stepIndex < manifest.steps.length - 1) {
-    const nextStep = manifest.steps[stepIndex + 1];
-    if (nextStep.state === "queued") {
-      nextStep.state = "waiting"; // ready to launch
-    }
-  }
-
-  return manifest;
-}
-
-/**
- * Get the chain's overall status from individual step states.
- *
- * @param {DispatchManifest} manifest
- * @returns {{status: string, completedSteps: number, totalSteps: number, currentStep: Object|null, blockedSteps: Object[]}}
- */
-export function getChainStatus(manifest) {
-  const completed = manifest.steps.filter(s => s.state === "completed").length;
-  const blocked = manifest.steps.filter(s => s.state === "blocked");
-  const failed = manifest.steps.filter(s => s.state === "failed");
-  const running = manifest.steps.find(s => s.state === "running");
-  const waiting = manifest.steps.find(s => s.state === "waiting");
-
-  let status;
-  if (failed.length > 0) status = "failed";
-  else if (blocked.length > 0) status = "blocked";
-  else if (completed === manifest.steps.length) status = "completed";
-  else if (running) status = "running";
-  else if (waiting) status = "ready";
-  else status = "queued";
-
-  return {
-    status,
-    completedSteps: completed,
-    totalSteps: manifest.steps.length,
-    currentStep: running || waiting || null,
-    blockedSteps: blocked,
-    failedSteps: failed,
-  };
-}
-
-// ── Escalation packet generator ───────────────────────────────────────────────
-
-/**
- * Generate an escalation packet when a step is blocked or rejected.
- *
- * @param {Object} step - The blocked/rejected step
- * @param {string} reason - Block/rejection reason
- * @param {string} verdict - "blocked" or "reject"
- * @returns {Object} Escalation packet ready for routing
- */
-export function generateEscalationPacket(step, reason, verdict) {
-  const escalation = verdict === "blocked"
-    ? resolveBlocked(reason)
-    : resolveRejected(reason, step.role);
-
-  return {
-    type: "escalation",
-    sourceStep: step.packetId,
-    sourceRole: step.role,
-    verdict,
-    reason,
-    escalation,
-    packet: {
-      title: `Escalation: ${step.role} ${verdict} — ${reason.slice(0, 80)}`,
-      assignedRole: escalation.targetRole,
-      recovery: escalation.recovery,
-      requiredArtifact: escalation.requiredArtifact,
-      context: escalation.handoffContext || escalation.reason,
-      sourcePacketFile: step.packetId,
-    },
-    generatedAt: new Date().toISOString(),
-  };
-}
-
-// ── Manifest persistence ──────────────────────────────────────────────────────
-
-/**
- * Save a dispatch manifest to disk.
- *
- * @param {DispatchManifest} manifest
- * @param {string} outputDir - Directory to write to
- * @returns {string} Path to the saved manifest
- */
-export function saveManifest(manifest, outputDir) {
-  mkdirSync(outputDir, { recursive: true });
-  const path = join(outputDir, `${manifest.runId}.dispatch.json`);
-  writeFileSync(path, JSON.stringify(manifest, null, 2));
-  return path;
-}
-
-/**
- * Load a dispatch manifest from disk.
- *
- * @param {string} path
- * @returns {DispatchManifest}
- */
-export function loadManifest(path) {
-  return JSON.parse(readFileSync(path, "utf-8"));
-}
-
-// ── Exports for integration ───────────────────────────────────────────────────
-
-export { TOOL_PROFILES, DEFAULTS };
+/**
+ * Runtime dispatch engine.
+ *
+ * Turns a staffed, validated chain into an executable dispatch manifest
+ * that multi-claude can consume. Role-OS owns staffing/routing/evidence;
+ * multi-claude owns execution.
+ *
+ * This module:
+ * 1. Maps roles → role configs (tool profiles, system prompts, budgets)
+ * 2. Tracks execution state (queued → running → completed/failed/blocked)
+ * 3. Collects outputs back into the packet system
+ * 4. Generates escalation packets for blocked/rejected work
+ * 5. Closes chains to final completion
+ */
+
+import { existsSync, mkdirSync, writeFileSync, readFileSync } from "node:fs";
+import { join, resolve } from "node:path";
+import { resolveBlocked, resolveRejected } from "./escalation.mjs";
+
+// ── Tool profiles per role ────────────────────────────────────────────────────
+// What tools each role is allowed to use. Sandboxing by contract.
+
+const TOOL_PROFILES = {
+  // Core
+  "Orchestrator":         ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
+  "Product Strategist":   ["Read", "Glob", "Grep", "Write"],
+  "Critic Reviewer":      ["Read", "Glob", "Grep", "Bash"],
+
+  // Design
+  "UI Designer":          ["Read", "Glob", "Grep", "Write", "Edit"],
+  "Brand Guardian":       ["Read", "Glob", "Grep"],
+
+  // Engineering
+  "Backend Engineer":     ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
+  "Frontend Developer":   ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
+  "Test Engineer":        ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
+  "Performance Engineer": ["Read", "Glob", "Grep", "Bash"],
+  "Refactor Engineer":    ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
+  "Security Reviewer":    ["Read", "Glob", "Grep", "Bash"],
+  "Dependency Auditor":   ["Read", "Glob", "Grep", "Bash"],
+
+  // Treatment
+  "Repo Researcher":      ["Read", "Glob", "Grep", "Bash"],
+  "Repo Translator":      ["Read", "Glob", "Grep", "Write", "Edit"],
+  "Docs Architect":       ["Read", "Glob", "Grep", "Write", "Edit"],
+  "Metadata Curator":     ["Read", "Glob", "Grep", "Write", "Edit"],
+  "Coverage Auditor":     ["Read", "Glob", "Grep", "Bash"],
+  "Deployment Verifier":  ["Read", "Glob", "Grep", "Bash"],
+  "Release Engineer":     ["Read", "Glob", "Grep", "Bash", "Write", "Edit"],
+
+  // Growth / Marketing
+  "Launch Strategist":    ["Read", "Glob", "Grep", "Write"],
+  "Content Strategist":   ["Read", "Glob", "Grep", "Write"],
+  "Community Manager":    ["Read", "Glob", "Grep", "Write"],
+  "Support Triage Lead":  ["Read", "Glob", "Grep", "Write"],
+  "Launch Copywriter":    ["Read", "Glob", "Grep", "Write", "Edit"],
+
+  // Product
+  "Feedback Synthesizer": ["Read", "Glob", "Grep"],
+  "Roadmap Prioritizer":  ["Read", "Glob", "Grep", "Write"],
+  "Spec Writer":          ["Read", "Glob", "Grep", "Write", "Edit"],
+
+  // Research
+  "UX Researcher":        ["Read", "Glob", "Grep"],
+  "Competitive Analyst":  ["Read", "Glob", "Grep"],
+  "Trend Researcher":     ["Read", "Glob", "Grep"],
+  "User Interview Synthesizer": ["Read", "Glob", "Grep"],
+
+  // Brainstorm
+  "Context Scout":        ["Read", "Glob", "Grep"],
+  "User Value Scout":     ["Read", "Glob", "Grep"],
+  "Creative Leap Scout":  ["Read", "Glob", "Grep"],
+  "Normalizer":           ["Read", "Glob", "Grep"],
+  "Synthesizer":          ["Read", "Glob", "Grep", "Write"],
+  "Product Expander":     ["Read", "Glob", "Grep", "Write"],
+  "Judge":                ["Read", "Glob", "Grep"],
+  "Mechanics Scout":      ["Read", "Glob", "Grep"],
+  "Market Scout":         ["Read", "Glob", "Grep"],
+  "Contrarian Scout":     ["Read", "Glob", "Grep"],
+  "Feasibility Scout":    ["Read", "Glob", "Grep"],
+  "Quality Bar Scout":    ["Read", "Glob", "Grep"],
+  "Scenario Expander":    ["Read", "Glob", "Grep", "Write"],
+  "Moat Expander":        ["Read", "Glob", "Grep", "Write"],
+
+  // Brainstorm v0.3 analysts
+  "Context Analyst":      ["Read", "Glob", "Grep"],
+  "User Value Analyst":   ["Read", "Glob", "Grep"],
+  "Mechanics Analyst":    ["Read", "Glob", "Grep"],
+  "Positioning Analyst":  ["Read", "Glob", "Grep"],
+  "Contrarian Analyst":   ["Read", "Glob", "Grep"],
+};
+
+// ── Default role config ─────────────────────────────────────────────────────
+
+const DEFAULTS = {
+  model: "sonnet",
+  maxTurns: 30,
+  maxBudgetUsd: 5.0,
+  timeoutMs: 10 * 60 * 1000, // 10 minutes
+};
+
+// ── Execution states ──────────────────────────────────────────────────────────
+
+export const EXEC_STATES = [
+  "queued",         // in chain, not yet launched
+  "running",        // role session active
+  "waiting",        // waiting for upstream handoff
+  "blocked",        // hit a block condition
+  "needs-review",   // completed work, awaiting verdict
+  "completed",      // approved and done
+  "failed",         // session error or timeout
+  "interrupted",    // manually stopped
+];
+
+// ── System prompt builder ─────────────────────────────────────────────────────
+
+function buildRolePrompt(roleName, packetContent, chainContext) {
+  return `You are operating as ${roleName} in a Role-OS managed chain.
+
+## Your Role Contract
+Follow the ${roleName} contract exactly. Your quality bar, inputs, outputs, and escalation rules are defined in .claude/agents/.
+
+## Current Packet
+${packetContent}
+
+## Chain Context
+You are step ${chainContext.stepNumber} of ${chainContext.totalSteps} in this chain.
+${chainContext.previousRole ? `Previous role: ${chainContext.previousRole} (${chainContext.previousStatus})` : "You are the first role in this chain."}
+${chainContext.nextRole ? `Next role: ${chainContext.nextRole}` : "You are the last role before Critic review."}
+
+## Handoff Requirements
+When you finish, produce a structured handoff:
+1. Summary of what you did
+2. Artifacts produced (files, changes, docs)
+3. Open questions or risks for the next role
+4. Evidence items for your verdict (kind, reference, claim, status)
+
+## Stop Conditions
+- If you cannot proceed: output a BLOCKED status with a clear reason
+- If the upstream handoff is insufficient: output a BLOCKED status explaining what's missing
+- Do not silently skip work — surface every gap explicitly
+`;
+}
+
+// ── Dispatch manifest builder ─────────────────────────────────────────────────
+
+/**
+ * Build a dispatch manifest from a routed chain.
+ * This is the contract between Role-OS (planning) and multi-claude (execution).
+ *
+ * @param {Object} options
+ * @param {string} options.packetFile - Path to the packet markdown
+ * @param {string} options.packetContent - Packet markdown content
+ * @param {Array<{role: {name: string, pack: string}}>} options.chainRoles - Assembled chain
+ * @param {string} options.cwd - Working directory for execution
+ * @param {Object} [options.overrides] - Per-role config overrides
+ * @returns {DispatchManifest}
+ */
+export function buildDispatchManifest({ packetFile, packetContent, chainRoles, cwd, overrides = {} }) {
+  const runId = `run-${Date.now()}`;
+  const steps = [];
+
+  for (let i = 0; i < chainRoles.length; i++) {
+    const { role } = chainRoles[i];
+    const roleOverrides = overrides[role.name] || {};
+
+    const chainContext = {
+      stepNumber: i + 1,
+      totalSteps: chainRoles.length,
+      previousRole: i > 0 ? chainRoles[i - 1].role.name : null,
+      previousStatus: i > 0 ? "pending" : null,
+      nextRole: i < chainRoles.length - 1 ? chainRoles[i + 1].role.name : null,
+    };
+
+    steps.push({
+      stepIndex: i,
+      packetId: `${runId}-step-${i}`,
+      role: role.name,
+      pack: role.pack,
+      tools: TOOL_PROFILES[role.name] || ["Read", "Glob", "Grep"],
+      systemPrompt: buildRolePrompt(role.name, packetContent, chainContext),
+      model: roleOverrides.model || DEFAULTS.model,
+      maxTurns: roleOverrides.maxTurns || DEFAULTS.maxTurns,
+      maxBudgetUsd: roleOverrides.maxBudgetUsd || DEFAULTS.maxBudgetUsd,
+      timeoutMs: roleOverrides.timeoutMs || DEFAULTS.timeoutMs,
+      state: "queued",
+      dependsOn: i > 0 ? [`${runId}-step-${i - 1}`] : [],
+      handoff: {
+        from: i > 0 ? chainRoles[i - 1].role.name : null,
+        to: i < chainRoles.length - 1 ? chainRoles[i + 1].role.name : null,
+      },
+    });
+  }
+
+  return {
+    version: 1,
+    runId,
+    createdAt: new Date().toISOString(),
+    packetFile: resolve(packetFile),
+    cwd: resolve(cwd),
+    totalSteps: steps.length,
+    steps,
+  };
+}
+
+// ── Execution state tracker ───────────────────────────────────────────────────
+
+/**
+ * Update a step's execution state in the manifest.
+ *
+ * @param {DispatchManifest} manifest
+ * @param {number} stepIndex
+ * @param {string} newState - One of EXEC_STATES
+ * @param {Object} [result] - Execution result data
+ * @returns {DispatchManifest} Updated manifest (mutated in place)
+ */
+export function updateStepState(manifest, stepIndex, newState, result = {}) {
+  const step = manifest.steps[stepIndex];
+  if (!step) throw new Error(`Invalid step index: ${stepIndex}`);
+  if (!EXEC_STATES.includes(newState)) throw new Error(`Invalid state: ${newState}`);
+
+  step.state = newState;
+  step.result = {
+    ...step.result,
+    ...result,
+    updatedAt: new Date().toISOString(),
+  };
+
+  // Auto-advance: if this step completed, mark next step as ready
+  if (newState === "completed" && stepIndex < manifest.steps.length - 1) {
+    const nextStep = manifest.steps[stepIndex + 1];
+    if (nextStep.state === "queued") {
+      nextStep.state = "waiting"; // ready to launch
+    }
+  }
+
+  return manifest;
+}
+
+/**
+ * Get the chain's overall status from individual step states.
+ *
+ * @param {DispatchManifest} manifest
+ * @returns {{status: string, completedSteps: number, totalSteps: number, currentStep: Object|null, blockedSteps: Object[]}}
+ */
+export function getChainStatus(manifest) {
+  const completed = manifest.steps.filter(s => s.state === "completed").length;
+  const blocked = manifest.steps.filter(s => s.state === "blocked");
+  const failed = manifest.steps.filter(s => s.state === "failed");
+  const running = manifest.steps.find(s => s.state === "running");
+  const waiting = manifest.steps.find(s => s.state === "waiting");
+
+  let status;
+  if (failed.length > 0) status = "failed";
+  else if (blocked.length > 0) status = "blocked";
+  else if (completed === manifest.steps.length) status = "completed";
+  else if (running) status = "running";
+  else if (waiting) status = "ready";
+  else status = "queued";
+
+  return {
+    status,
+    completedSteps: completed,
+    totalSteps: manifest.steps.length,
+    currentStep: running || waiting || null,
+    blockedSteps: blocked,
+    failedSteps: failed,
+  };
+}
+
+// ── Escalation packet generator ───────────────────────────────────────────────
+
+/**
+ * Generate an escalation packet when a step is blocked or rejected.
+ *
+ * @param {Object} step - The blocked/rejected step
+ * @param {string} reason - Block/rejection reason
+ * @param {string} verdict - "blocked" or "reject"
+ * @returns {Object} Escalation packet ready for routing
+ */
+export function generateEscalationPacket(step, reason, verdict) {
+  const escalation = verdict === "blocked"
+    ? resolveBlocked(reason)
+    : resolveRejected(reason, step.role);
+
+  return {
+    type: "escalation",
+    sourceStep: step.packetId,
+    sourceRole: step.role,
+    verdict,
+    reason,
+    escalation,
+    packet: {
+      title: `Escalation: ${step.role} ${verdict} — ${reason.slice(0, 80)}`,
+      assignedRole: escalation.targetRole,
+      recovery: escalation.recovery,
+      requiredArtifact: escalation.requiredArtifact,
+      context: escalation.handoffContext || escalation.reason,
+      sourcePacketFile: step.packetId,
+    },
+    generatedAt: new Date().toISOString(),
+  };
+}
+
+// ── Manifest persistence ──────────────────────────────────────────────────────
+
+/**
+ * Save a dispatch manifest to disk.
+ *
+ * @param {DispatchManifest} manifest
+ * @param {string} outputDir - Directory to write to
+ * @returns {string} Path to the saved manifest
+ */
+export function saveManifest(manifest, outputDir) {
+  mkdirSync(outputDir, { recursive: true });
+  const path = join(outputDir, `${manifest.runId}.dispatch.json`);
+  writeFileSync(path, JSON.stringify(manifest, null, 2));
+  return path;
+}
+
+/**
+ * Load a dispatch manifest from disk.
+ *
+ * @param {string} path
+ * @returns {DispatchManifest}
+ */
+export function loadManifest(path) {
+  return JSON.parse(readFileSync(path, "utf-8"));
+}
+
+// ── Exports for integration ───────────────────────────────────────────────────
+
+export { TOOL_PROFILES, DEFAULTS };