forgesmith 0.6.1 → 0.7.0

package/dist/index.mjs

@@ -4,12 +4,12 @@ import { randomUUID } from 'crypto';
 function buildSystemPrompt() {
   return `You are a technical writer and developer-relations expert. You generate clear, accurate release notes from structured code-intelligence data. Write only the release notes \u2014 no preamble, no meta-commentary.`;
 }
-function buildUserPrompt(data, opts) {
+function buildUserPrompt(data, opts, amberContext, greenContext) {
   const tone = opts.tone ?? "professional";
   const length = opts.length ?? "medium";
   const format = opts.format ?? "markdown";
   const lengthGuide = { short: "2-3 paragraphs or bullet groups", medium: "4-6 sections", long: "comprehensive, 6+ sections with details" }[length];
-  const toneGuide = { professional: "formal, clear, business-appropriate", casual: "friendly, approachable, conversational", technical: "precise, implementation-focused, developer-centric" }[tone];
+  const toneGuide = { professional: "formal, clear, business-appropriate", casual: "friendly, approachable, conversational", technical: "precise, implementation-focused, developer-centric", executive: "high-level, strategic, board-ready summary" }[tone] ?? "formal, clear, business-appropriate";
   const hasSessions = (data.sessions?.length ?? 0) > 0;
   const hasRecs = (data.recommendations?.length ?? 0) > 0;
   const hasInsights = (data.insights?.length ?? 0) > 0;
@@ -45,11 +45,38 @@ function buildUserPrompt(data, opts) {
     }
     lines.push(``);
   }
+  if (amberContext && amberContext.capabilities.length > 0) {
+    lines.push(`## Business Capabilities (AMBER layer)`);
+    lines.push(amberContext.summary);
+    const criticalCaps = amberContext.capabilities.filter(
+      (c) => c.criticality === "critical" || c.criticality === "high"
+    );
+    if (criticalCaps.length > 0) {
+      lines.push(
+        `This release may touch these capabilities: ${criticalCaps.map((c) => c.name).join(", ")}.`
+      );
+    }
+    lines.push(``);
+  }
+  if (greenContext) {
+    if (greenContext.coherenceScore !== null) {
+      lines.push(`## Architecture Health (GREEN layer)`);
+      lines.push(
+        `Coherence score: ${greenContext.coherenceScore}/100 (grade ${greenContext.coherenceGrade ?? "?"}).`
+      );
+      if (greenContext.topRisks.length > 0) {
+        lines.push(
+          `Top risks: ${greenContext.topRisks.slice(0, 3).map((r) => `${r.name} \u2014 ${r.reason}`).join("; ")}.`
+        );
+      }
+      lines.push(``);
+    }
+  }
   return lines.join("\n");
 }
-async function generateReleaseNotes(data, opts, provider) {
+async function generateReleaseNotes(data, opts, provider, amberContext, greenContext) {
   const systemPrompt = buildSystemPrompt();
-  const userPrompt = buildUserPrompt(data, opts);
+  const userPrompt = buildUserPrompt(data, opts, amberContext, greenContext);
   const response = await provider.complete({
     systemPrompt,
     messages: [{ role: "user", content: userPrompt }],
@@ -69,13 +96,16 @@ async function generateReleaseNotes(data, opts, provider) {
 var NO_DATA_MSG = "No Blueprint data available. Run prism scan first.";
 function buildSystemPrompt2(tone) {
   const toneMap = {
-    technical: "You are a senior software architect who writes precise, technical documentation for engineering teams.",
+    professional: "You are a senior software architect who writes clear, professional documentation suitable for all engineering audiences.",
     casual: "You are a friendly developer advocate who explains codebases in an approachable, conversational style.",
+    technical: "You are a senior software architect who writes precise, technical documentation for engineering teams.",
+    executive: "You are a CTO-level advisor who explains codebase architecture at a high strategic level for leadership audiences.",
+    // Legacy aliases kept for backwards compatibility
     onboarding: "You are an experienced engineering mentor who helps new developers quickly understand a codebase."
   };
-  return `${toneMap[tone] ?? toneMap.technical} Write only the requested document \u2014 no preamble, no meta-commentary.`;
+  return `${toneMap[tone] ?? toneMap.professional} Write only the requested document \u2014 no preamble, no meta-commentary.`;
 }
-function buildUserPrompt2(blueprint, opts) {
+function buildUserPrompt2(blueprint, opts, amberContext) {
   const tone = opts.tone ?? "technical";
   const length = opts.length ?? "long";
   const format = opts.format ?? "markdown";
@@ -103,9 +133,28 @@ function buildUserPrompt2(blueprint, opts) {
     ``,
     `Cover: architecture overview, key layers/folders, entry points, important relationships, and what a new developer should understand first.`
   ];
+  if (amberContext && amberContext.capabilities.length > 0) {
+    lines.push(``);
+    lines.push(`## Business Capabilities (AMBER layer)`);
+    lines.push(amberContext.summary);
+    lines.push(``);
+    const sorted = amberContext.capabilities.slice().sort((a, b) => {
+      const rank = { critical: 0, high: 1, medium: 2, low: 3 };
+      return (rank[a.criticality] ?? 4) - (rank[b.criticality] ?? 4);
+    });
+    for (const cap of sorted) {
+      const tags = [cap.criticality, cap.lifecycle].filter(Boolean).join(", ");
+      const fileCount = cap.files.length;
+      lines.push(
+        `- **${cap.name}** (${cap.id}) [${tags}]${fileCount > 0 ? ` \u2014 ${fileCount} file${fileCount === 1 ? "" : "s"}` : ""}`
+      );
+      if (cap.description) lines.push(`  ${cap.description}`);
+    }
+    lines.push(`Include a "Business Capabilities" section in the walkthrough mapping these capabilities to the codebase.`);
+  }
   return lines.join("\n");
 }
-async function generateArchitectureWalkthrough(blueprint, opts, provider) {
+async function generateArchitectureWalkthrough(blueprint, opts, provider, amberContext) {
   if (!blueprint) {
     return {
       text: NO_DATA_MSG,
@@ -114,7 +163,7 @@ async function generateArchitectureWalkthrough(blueprint, opts, provider) {
   }
   const response = await provider.complete({
     systemPrompt: buildSystemPrompt2(opts.tone ?? "technical"),
-    messages: [{ role: "user", content: buildUserPrompt2(blueprint, opts) }],
+    messages: [{ role: "user", content: buildUserPrompt2(blueprint, opts, amberContext) }],
     maxTokens: 4096
   });
   return {
@@ -189,11 +238,15 @@ async function generateChangesSince(current, previous, opts, provider) {
 var NO_DATA_MSG3 = "No Blueprint data available. Run prism scan first.";
 function buildSystemPrompt4(tone) {
   const toneMap = {
+    professional: "You are a documentation engineer writing a structured, professional onboarding reference for engineering teams.",
+    casual: "You are a helpful senior developer writing a welcoming onboarding guide for new team members. Be warm, encouraging, and practical.",
+    technical: "You are a senior engineer writing a technical onboarding document focused on implementation details and system internals.",
+    executive: "You are a technical lead writing a concise onboarding overview suitable for both technical and leadership audiences.",
+    // Legacy aliases kept for backwards compatibility
     friendly: "You are a helpful senior developer writing a welcoming onboarding guide for new team members. Be warm, encouraging, and practical.",
-    formal: "You are a documentation engineer writing a structured onboarding reference for enterprise engineering teams.",
-    technical: "You are a senior engineer writing a technical onboarding document focused on implementation details and system internals."
+    formal: "You are a documentation engineer writing a structured onboarding reference for enterprise engineering teams."
   };
-  return `${toneMap[tone] ?? toneMap.friendly} Write only the document \u2014 no preamble, no meta-commentary.`;
+  return `${toneMap[tone] ?? toneMap.professional} Write only the document \u2014 no preamble, no meta-commentary.`;
 }
 function buildUserPrompt4(blueprint, opts) {
   const tone = opts.tone ?? "friendly";
@@ -254,11 +307,14 @@ async function generateOnboardingDoc(blueprint, opts, provider) {
 var NO_DATA_MSG4 = "No Blueprint data available. Run prism scan first.";
 function buildSystemPrompt5(tone) {
   const toneMap = {
-    analytical: "You are a software architecture consultant who produces rigorous, evidence-based refactoring reports with clear prioritization.",
+    professional: "You are a software architecture consultant who produces rigorous, evidence-based refactoring reports with clear prioritization.",
     casual: "You are a senior developer who gives honest, practical refactoring advice based on code structure data.",
-    executive: "You are a CTO-level advisor who summarizes architectural debt and refactoring priorities for engineering leadership."
+    technical: "You are a senior engineer who produces detailed, implementation-focused refactoring analysis with concrete code-level recommendations.",
+    executive: "You are a CTO-level advisor who summarizes architectural debt and refactoring priorities for engineering leadership.",
+    // Legacy aliases kept for backwards compatibility
+    analytical: "You are a software architecture consultant who produces rigorous, evidence-based refactoring reports with clear prioritization."
   };
-  return `${toneMap[tone] ?? toneMap.analytical} Write only the report \u2014 no preamble, no meta-commentary.`;
+  return `${toneMap[tone] ?? toneMap.professional} Write only the report \u2014 no preamble, no meta-commentary.`;
 }
 function detectImportCycles(edges) {
   const graph = /* @__PURE__ */ new Map();
@@ -266,158 +322,2524 @@ function detectImportCycles(edges) {
     if (!graph.has(e.from)) graph.set(e.from, /* @__PURE__ */ new Set());
     graph.get(e.from).add(e.to);
   }
-  const cycles = [];
-  const visited = /* @__PURE__ */ new Set();
-  const stack = /* @__PURE__ */ new Set();
-  function dfs(node, path) {
-    if (cycles.length >= 5) return;
-    if (stack.has(node)) {
-      const cycleStart = path.indexOf(node);
-      if (cycleStart !== -1) {
-        cycles.push(path.slice(cycleStart).join(" \u2192 ") + " \u2192 " + node);
+  const cycles = [];
+  const visited = /* @__PURE__ */ new Set();
+  const stack = /* @__PURE__ */ new Set();
+  function dfs(node, path) {
+    if (cycles.length >= 5) return;
+    if (stack.has(node)) {
+      const cycleStart = path.indexOf(node);
+      if (cycleStart !== -1) {
+        cycles.push(path.slice(cycleStart).join(" \u2192 ") + " \u2192 " + node);
+      }
+      return;
+    }
+    if (visited.has(node)) return;
+    visited.add(node);
+    stack.add(node);
+    for (const neighbor of graph.get(node) ?? []) {
+      dfs(neighbor, [...path, node]);
+    }
+    stack.delete(node);
+  }
+  for (const node of graph.keys()) {
+    if (!visited.has(node)) dfs(node, []);
+  }
+  return cycles;
+}
+function buildUserPrompt5(blueprint, opts, amberContext, greenContext) {
+  const tone = opts.tone ?? "analytical";
+  const length = opts.length ?? "medium";
+  const format = opts.format ?? "markdown";
+  const hotFiles = blueprint.files.slice().sort((a, b) => (b.importedByCount ?? 0) + (b.importCount ?? 0) - ((a.importedByCount ?? 0) + (a.importCount ?? 0))).slice(0, 10);
+  const highFanOut = blueprint.files.slice().sort((a, b) => (b.importCount ?? 0) - (a.importCount ?? 0)).slice(0, 8);
+  const importCycles = detectImportCycles(blueprint.edges);
+  const lines = [
+    `Generate a refactoring report in ${format} format for the codebase at "${blueprint.targetPath}".`,
+    `Tone: ${tone}. Length: ${length}.`,
+    ``,
+    `## Architecture Metrics`,
+    `- Total files: ${blueprint.stats.totalFiles}`,
+    `- Dependency edges: ${blueprint.stats.runtimeEdges}`,
+    `- Categories: ${JSON.stringify(blueprint.categories)}`,
+    ``,
+    `## High-Coupling Hot Spots (high total connections)`,
+    ...hotFiles.map((f) => `- ${f.path} \u2014 in: ${f.importedByCount ?? 0}, out: ${f.importCount ?? 0}, lines: ${f.lineCount ?? "?"}`),
+    ``,
+    `## High Fan-Out Files (many outgoing dependencies)`,
+    ...highFanOut.map((f) => `- ${f.path} \u2014 imports: ${f.importCount ?? 0}`),
+    ``,
+    `## Import Cycles Detected (${importCycles.length})`,
+    ...importCycles.length > 0 ? importCycles.map((c) => `- ${c}`) : ["No cycles detected in sampled edges"],
+    ``,
+    `Identify refactoring priorities: coupling issues, over-large files, circular dependencies, layer violations. Suggest concrete refactoring actions with rationale.`
+  ];
+  if (amberContext && amberContext.capabilities.length > 0) {
+    lines.push(``);
+    lines.push(`## Business Capabilities with Documentation Drift (AMBER layer)`);
+    lines.push(amberContext.summary);
+    const drifted = amberContext.capabilities.filter((c) => c.driftCount > 0);
+    if (drifted.length > 0) {
+      lines.push(`Capabilities with drift (files changed without updating @amber-doc):`);
+      for (const cap of drifted) {
+        lines.push(
+          `- **${cap.name}** (${cap.id}, ${cap.criticality}): ${cap.driftCount} file${cap.driftCount === 1 ? "" : "s"} drifted`
+        );
+      }
+    }
+  }
+  if (greenContext) {
+    lines.push(``);
+    lines.push(`## Cross-Layer Risks (GREEN layer)`);
+    lines.push(greenContext.summary);
+    if (greenContext.topRisks.length > 0) {
+      lines.push(`Top architectural risks:`);
+      for (const risk of greenContext.topRisks) {
+        lines.push(
+          `- **${risk.name}** (impact: ${risk.impact}): ${risk.reason}`
+        );
+      }
+    }
+    if (greenContext.insights.length > 0) {
+      lines.push(`Cross-layer insights:`);
+      for (const insight of greenContext.insights.slice(0, 5)) {
+        lines.push(`- **${insight.title}**: ${insight.body}`);
+      }
+    }
+  }
+  return lines.join("\n");
+}
+async function generateRefactoringReport(blueprint, opts, provider, amberContext, greenContext) {
+  if (!blueprint) {
+    return {
+      text: NO_DATA_MSG4,
+      metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: 0, generator: "forgesmith" }
+    };
+  }
+  const response = await provider.complete({
+    systemPrompt: buildSystemPrompt5(opts.tone ?? "analytical"),
+    messages: [{ role: "user", content: buildUserPrompt5(blueprint, opts, amberContext, greenContext) }],
+    maxTokens: 3072
+  });
+  return {
+    text: response.content,
+    metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: response.usedTokens, generator: "forgesmith" }
+  };
+}
+
+// src/generators/askDrivenAsset.ts
+var NO_DATA_MSG5 = "No Blueprint data available. Run prism scan first.";
+var FORMAT_GUIDES = {
+  markdown: { name: "Markdown document", structure: "Use headers, bullet lists, and code blocks where appropriate.", maxTokens: 3072 },
+  blog: { name: "blog post", structure: "Write with an engaging intro, clear sections, a conclusion, and a call-to-action.", maxTokens: 3072 },
+  social: { name: "social media post", structure: "Write concise, punchy content suitable for Twitter/LinkedIn. Max 280 characters for Twitter mode.", maxTokens: 512 },
+  email: { name: "email", structure: "Use Subject:, greeting, body paragraphs, and a sign-off.", maxTokens: 1024 },
+  slack: { name: "Slack message", structure: "Keep it conversational, use *bold* for emphasis, bullet points for lists. Max 3 paragraphs.", maxTokens: 512 },
+  slide: { name: "presentation outline", structure: "Structure as slide titles with 3-5 bullet points each. Include a title slide and summary slide.", maxTokens: 2048 }
+};
+function buildSystemPrompt6(format, tone) {
+  const toneMap = {
+    professional: "You are a professional technical writer and developer advocate.",
+    casual: "You are a friendly engineering blogger who writes in an approachable, conversational style.",
+    technical: "You are a senior software engineer writing precise, implementation-focused content.",
+    executive: "You are a VP of Engineering writing high-level, business-value-focused content for leadership."
+  };
+  const guide = FORMAT_GUIDES[format];
+  return `${toneMap[tone] ?? toneMap.professional} Generate a ${guide.name} based on the user's question and the provided codebase architecture context. ${guide.structure} Write only the requested content \u2014 no preamble, no meta-commentary.`;
+}
+function buildUserPrompt6(blueprint, question, opts) {
+  const format = opts.format ?? "markdown";
+  const tone = opts.tone ?? "professional";
+  const length = opts.length ?? "medium";
+  const guide = FORMAT_GUIDES[format];
+  const lengthGuide = {
+    short: "Keep it concise \u2014 1-2 paragraphs or equivalent.",
+    medium: "Medium length \u2014 3-5 paragraphs or equivalent.",
+    long: "Detailed and comprehensive \u2014 cover the topic thoroughly."
+  }[length];
+  const topFiles = blueprint.files.slice().sort((a, b) => (b.importedByCount ?? 0) - (a.importedByCount ?? 0)).slice(0, 12);
+  const lines = [
+    `## User's Question`,
+    question,
+    ``,
+    `## Output Requirements`,
+    `- Format: ${guide.name}`,
+    `- Tone: ${tone}`,
+    `- Length: ${length} (${lengthGuide})`,
+    ``,
+    `## Codebase Architecture Context`,
+    `Target: ${blueprint.targetPath}`,
+    `Total files: ${blueprint.stats.totalFiles} | Dependency edges: ${blueprint.stats.runtimeEdges}`,
+    `Categories: app=${blueprint.categories.app ?? 0}, components=${blueprint.categories.component ?? 0}, lib=${blueprint.categories.lib ?? 0}, hooks=${blueprint.categories.hook ?? 0}`,
+    ``,
+    `Key files (by usage):`,
+    ...topFiles.map((f) => `- ${f.path} [${f.category ?? "?"}] \u2014 imported by ${f.importedByCount ?? 0} files, ${f.lineCount ?? "?"} lines`)
+  ];
+  if (blueprint.edges.length > 0) {
+    const edgeSample = blueprint.edges.slice(0, 15);
+    lines.push(``, `Dependency edges (sample):`, ...edgeSample.map((e) => `- ${e.from} \u2192 ${e.to}`));
+  }
+  lines.push(``, `Answer the user's question using the architecture context above. Generate the ${guide.name} now.`);
+  return lines.join("\n");
+}
+async function generateAskDrivenAsset(blueprint, question, opts, provider) {
+  if (!blueprint) {
+    return {
+      text: NO_DATA_MSG5,
+      metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: 0, generator: "forgesmith" }
+    };
+  }
+  if (!question || !question.trim()) {
+    return {
+      text: "No question provided. Please ask something about your codebase.",
+      metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: 0, generator: "forgesmith" }
+    };
+  }
+  const format = opts.format ?? "markdown";
+  const tone = opts.tone ?? "professional";
+  const maxTokens = FORMAT_GUIDES[format].maxTokens;
+  const response = await provider.complete({
+    systemPrompt: buildSystemPrompt6(format, tone),
+    messages: [{ role: "user", content: buildUserPrompt6(blueprint, question, opts) }],
+    maxTokens
+  });
+  return {
+    text: response.content,
+    metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: response.usedTokens, generator: "forgesmith" }
+  };
+}
+
+// src/generators/generatePresentation.ts
+function buildContext(blueprint, opts) {
+  const topFiles = blueprint.files.slice().sort((a, b) => (b.importedByCount ?? 0) - (a.importedByCount ?? 0)).slice(0, 10);
+  const projectName = opts.projectName || blueprint.targetPath.split("/").filter(Boolean).pop() || "Project";
+  const totalLoc = blueprint.files.reduce(
+    (sum, f) => sum + (f.lineCount ?? 0),
+    0
+  );
+  const lines = [
+    `## Codebase Context`,
+    `Project: ${projectName}`,
+    `Path: ${blueprint.targetPath}`,
+    `Total files: ${blueprint.stats.totalFiles}`,
+    `Total lines of code: ${totalLoc}`,
+    `Dependency edges: ${blueprint.stats.runtimeEdges}`,
+    `Categories: app=${blueprint.categories.app ?? 0}, components=${blueprint.categories.component ?? 0}, lib=${blueprint.categories.lib ?? 0}, hooks=${blueprint.categories.hook ?? 0}`,
+    ``,
+    `## Most-imported files`,
+    ...topFiles.map(
+      (f) => `- ${f.path} [${f.category ?? "unknown"}] \u2014 importedBy: ${f.importedByCount ?? 0}, lines: ${f.lineCount ?? "?"}`
+    )
+  ];
+  if (blueprint.edges.length > 0) {
+    const edgeSample = blueprint.edges.slice(0, 10);
+    lines.push(
+      ``,
+      `## Dependency sample`,
+      ...edgeSample.map((e) => `- ${e.from} \u2192 ${e.to}${e.type ? ` (${e.type})` : ""}`)
+    );
+  }
+  return lines.join("\n");
+}
+function planSlideTypes(opts) {
+  const { slideCount, focusArea, audience } = opts;
+  const focus = focusArea ?? "all";
+  const always = ["title", "closing"];
+  let middle = [];
+  if (focus === "architecture") {
+    middle = ["executive", "architecture", "metrics", "risks", "recommendations"];
+  } else if (focus === "health") {
+    middle = ["executive", "health", "risks", "recommendations", "metrics"];
+  } else if (focus === "risks") {
+    middle = ["executive", "risks", "recommendations", "architecture", "health"];
+  } else {
+    if (audience === "executive") {
+      middle = ["executive", "health", "risks", "recommendations", "architecture", "metrics", "capabilities", "recommendations"];
+    } else if (audience === "technical") {
+      middle = ["architecture", "metrics", "risks", "health", "recommendations", "capabilities", "executive", "recommendations"];
+    } else {
+      middle = ["executive", "architecture", "health", "risks", "recommendations", "metrics", "capabilities", "executive"];
+    }
+  }
+  const targetMiddle = slideCount - 2;
+  const trimmed = middle.slice(0, targetMiddle);
+  const seen = /* @__PURE__ */ new Set();
+  const deduped = [];
+  for (const t of trimmed) {
+    if (!seen.has(t)) {
+      seen.add(t);
+      deduped.push(t);
+    }
+  }
+  return [always[0], ...deduped, always[1]];
+}
+function buildSystemPrompt7(opts) {
+  const toneMap = {
+    professional: "You are a senior engineering consultant who creates compelling, data-driven presentations for technical and business stakeholders.",
+    casual: "You are a developer advocate who creates friendly, approachable presentations that make architecture exciting.",
+    executive: "You are a VP of Engineering who creates concise, business-value-focused presentations for C-suite audiences."
+  };
+  return `${toneMap[opts.tone]} Generate ONLY valid JSON \u2014 no markdown fences, no prose, no commentary outside the JSON object.`;
+}
+function buildUserPrompt7(blueprint, opts, slideTypes) {
+  const projectName = opts.projectName || blueprint.targetPath.split("/").filter(Boolean).pop() || "Project";
+  const context = buildContext(blueprint, opts);
+  const slideTypeDescriptions = {
+    title: "Opening slide with project name, tagline, and date. Fields: title (project name), subtitle (tagline), speakerNotes",
+    executive: "Executive summary with 3 key findings. Fields: title, bullets (exactly 3 bullet strings), speakerNotes",
+    architecture: "Architecture overview. Fields: title, subtitle (layer summary), bullets (up to 5 key architecture facts), visualHint (describe a diagram), speakerNotes",
+    capabilities: "Capability map overview. Fields: title, bullets (up to 5 capability areas found), speakerNotes",
+    health: "Codebase health score. Fields: title, highlight (a letter grade A-F or numeric score), highlightLabel, bullets (up to 3 health indicators), speakerNotes",
+    risks: "Top risks. Fields: title, bullets (exactly 3 risks, each prefixed with High/Medium/Low impact), speakerNotes",
+    recommendations: "Action items. Fields: title, bullets (up to 5 concrete action items), speakerNotes",
+    metrics: "Key metrics. Fields: title, highlight (most important metric), highlightLabel, bullets (up to 5 metrics), speakerNotes",
+    closing: "Next steps and call to action. Fields: title, subtitle, bullets (up to 3 next steps), speakerNotes"
+  };
+  const slidePlan = slideTypes.map((t, i) => `  ${i + 1}. type="${t}" \u2014 ${slideTypeDescriptions[t]}`).join("\n");
+  return [
+    `Generate a ${opts.slideCount}-slide presentation deck for: ${projectName}`,
+    `Audience: ${opts.audience}. Tone: ${opts.tone}. Theme: ${opts.theme}.`,
+    ``,
+    context,
+    ``,
+    `## Slide Plan (${slideTypes.length} slides)`,
+    slidePlan,
+    ``,
+    `## Output Format`,
+    `Return a single JSON object with this exact shape:`,
+    `{`,
+    `  "title": "<deck title>",`,
+    `  "subtitle": "<deck subtitle>",`,
+    `  "slides": [`,
+    `    {`,
+    `      "type": "<one of the PresentationSlideType values>",`,
+    `      "title": "<slide title>",`,
+    `      "subtitle": "<optional subtitle>",`,
+    `      "bullets": ["<bullet 1>", "<bullet 2>"],`,
+    `      "highlight": "<optional large callout>",`,
+    `      "highlightLabel": "<optional label>",`,
+    `      "speakerNotes": "<presenter notes>",`,
+    `      "visualHint": "<optional visual description>"`,
+    `    }`,
+    `  ]`,
+    `}`,
+    ``,
+    `Rules:`,
+    `- Each slide in the output MUST match the type listed in the slide plan above.`,
+    `- bullets array: max 5 items, each \u2264 80 characters.`,
+    `- speakerNotes: 1-3 sentences a presenter would say.`,
+    `- Derive all content from the codebase context above \u2014 no invented facts.`,
+    `- Return ONLY valid JSON. No markdown. No prose outside the JSON.`
+  ].join("\n");
+}
+function buildFallback(blueprint, opts, slideTypes) {
+  const projectName = opts.projectName || blueprint.targetPath.split("/").filter(Boolean).pop() || "Project";
+  const totalLoc = blueprint.files.reduce(
+    (sum, f) => sum + (f.lineCount ?? 0),
+    0
+  );
+  const slideBuilders = {
+    title: () => ({
+      type: "title",
+      title: projectName,
+      subtitle: "Architecture & Code Intelligence Report",
+      speakerNotes: `Welcome. Today we cover the architecture of ${projectName}.`
+    }),
+    executive: () => ({
+      type: "executive",
+      title: "Executive Summary",
+      bullets: [
+        `${blueprint.stats.totalFiles} files analyzed across all layers`,
+        `${blueprint.stats.runtimeEdges} dependency edges mapped`,
+        `Architecture categories: app, components, lib, hooks`
+      ],
+      speakerNotes: "Three key findings from the codebase scan."
+    }),
+    architecture: () => ({
+      type: "architecture",
+      title: "Architecture Overview",
+      subtitle: `${blueprint.stats.totalFiles} files, ${blueprint.stats.runtimeEdges} edges`,
+      bullets: [
+        `App layer: ${blueprint.categories.app ?? 0} files`,
+        `Components: ${blueprint.categories.component ?? 0} files`,
+        `Libraries: ${blueprint.categories.lib ?? 0} files`,
+        `Hooks: ${blueprint.categories.hook ?? 0} files`
+      ],
+      visualHint: "Layered architecture diagram with dependency arrows",
+      speakerNotes: "The codebase is organized into four primary layers."
+    }),
+    capabilities: () => ({
+      type: "capabilities",
+      title: "Capability Map",
+      bullets: ["Core application logic", "Component layer", "Shared utilities", "Custom hooks"],
+      speakerNotes: "These are the key capability areas identified in the scan."
+    }),
+    health: () => ({
+      type: "health",
+      title: "Codebase Health",
+      highlight: "B",
+      highlightLabel: "Overall Grade",
+      bullets: [
+        `${blueprint.stats.totalFiles} total files`,
+        `${blueprint.stats.runtimeEdges} dependency edges`,
+        `Avg LOC per file: ${blueprint.stats.totalFiles > 0 ? Math.round(totalLoc / blueprint.stats.totalFiles) : 0}`
+      ],
+      speakerNotes: "The codebase shows good structural health with opportunities for improvement."
+    }),
+    risks: () => ({
+      type: "risks",
+      title: "Top Risks",
+      bullets: [
+        "High impact: Large files with many incoming dependencies",
+        "Medium impact: Missing test coverage in critical paths",
+        "Low impact: Inconsistent category classification"
+      ],
+      speakerNotes: "Three risks identified, prioritized by impact."
+    }),
+    recommendations: () => ({
+      type: "recommendations",
+      title: "Recommendations",
+      bullets: [
+        "Break down files with high importedByCount into smaller modules",
+        "Establish clear boundaries between app, lib, and component layers",
+        "Add test coverage to files with the most incoming dependencies",
+        "Review and resolve circular dependency patterns",
+        "Document architecture decisions for new team members"
+      ],
+      speakerNotes: "Five concrete action items, ordered by priority."
+    }),
+    metrics: () => ({
+      type: "metrics",
+      title: "Key Metrics",
+      highlight: String(blueprint.stats.totalFiles),
+      highlightLabel: "Total Files",
+      bullets: [
+        `Lines of code: ${totalLoc.toLocaleString()}`,
+        `Dependency edges: ${blueprint.stats.runtimeEdges}`,
+        `App files: ${blueprint.categories.app ?? 0}`,
+        `Components: ${blueprint.categories.component ?? 0}`,
+        `Libraries: ${blueprint.categories.lib ?? 0}`
+      ],
+      speakerNotes: "Key quantitative metrics from the Blueprint scan."
+    }),
+    closing: () => ({
+      type: "closing",
+      title: "Next Steps",
+      subtitle: "Start with the highest-impact recommendations",
+      bullets: [
+        "Schedule architecture review with the team",
+        "Prioritize the top 3 recommendations",
+        "Set up regular prism0x2A scans"
+      ],
+      speakerNotes: "Thank you. Questions?"
+    })
+  };
+  const slides = slideTypes.map((t) => slideBuilders[t]());
+  return {
+    title: `${projectName} \u2014 Architecture Deck`,
+    subtitle: `${opts.audience.charAt(0).toUpperCase() + opts.audience.slice(1)} presentation \xB7 ${(/* @__PURE__ */ new Date()).toLocaleDateString()}`,
+    slideCount: slides.length,
+    slides,
+    theme: opts.theme,
+    generatedAt: (/* @__PURE__ */ new Date()).toISOString()
+  };
+}
+function parseLlmResponse(raw, slideTypes, opts, blueprint) {
+  let text = raw.trim();
+  if (text.startsWith("```")) {
+    text = text.replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "").trim();
+  }
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  const rawSlides = Array.isArray(parsed.slides) ? parsed.slides : [];
+  if (rawSlides.length === 0) return null;
+  const slides = rawSlides.map((s, i) => {
+    const slide = s ?? {};
+    const inferredType = typeof slide.type === "string" && [
+      "title",
+      "executive",
+      "architecture",
+      "capabilities",
+      "health",
+      "risks",
+      "recommendations",
+      "metrics",
+      "closing"
+    ].includes(slide.type) ? slide.type : slideTypes[i] ?? "executive";
+    const bullets = Array.isArray(slide.bullets) ? slide.bullets.filter((b) => typeof b === "string").slice(0, 5) : void 0;
+    return {
+      type: inferredType,
+      title: typeof slide.title === "string" ? slide.title : `Slide ${i + 1}`,
+      ...typeof slide.subtitle === "string" && slide.subtitle ? { subtitle: slide.subtitle } : {},
+      ...bullets && bullets.length > 0 ? { bullets } : {},
+      ...typeof slide.highlight === "string" && slide.highlight ? { highlight: slide.highlight } : {},
+      ...typeof slide.highlightLabel === "string" && slide.highlightLabel ? { highlightLabel: slide.highlightLabel } : {},
+      ...typeof slide.speakerNotes === "string" && slide.speakerNotes ? { speakerNotes: slide.speakerNotes } : {},
+      ...typeof slide.visualHint === "string" && slide.visualHint ? { visualHint: slide.visualHint } : {}
+    };
+  });
+  const projectName = opts.projectName || blueprint.targetPath.split("/").filter(Boolean).pop() || "Project";
+  return {
+    title: typeof parsed.title === "string" && parsed.title ? parsed.title : `${projectName} \u2014 Architecture Deck`,
+    subtitle: typeof parsed.subtitle === "string" && parsed.subtitle ? parsed.subtitle : `${opts.audience} presentation`,
+    slideCount: slides.length,
+    slides,
+    theme: opts.theme,
+    generatedAt: (/* @__PURE__ */ new Date()).toISOString()
+  };
+}
+async function generatePresentation(opts) {
+  const { blueprint, llm } = opts;
+  const slideTypes = planSlideTypes(opts);
+  try {
+    const response = await llm.complete({
+      systemPrompt: buildSystemPrompt7(opts),
+      messages: [
+        {
+          role: "user",
+          content: buildUserPrompt7(blueprint, opts, slideTypes)
+        }
+      ],
+      maxTokens: 4096
+    });
+    const parsed = parseLlmResponse(response.content, slideTypes, opts, blueprint);
+    if (parsed) return parsed;
+  } catch {
+  }
+  return buildFallback(blueprint, opts, slideTypes);
+}
+
+// src/generators/generateComplianceDoc.ts
+var FRAMEWORK_SPECS = {
+  SOX: {
+    fullName: "Sarbanes-Oxley Act (SOX)",
+    focusAreas: [
+      "Internal controls over financial reporting (ICFR)",
+      "Access control and segregation of duties",
+      "Change management procedures",
+      "Audit trail completeness",
+      "IT general controls (ITGCs)"
+    ],
+    requiredSections: [
+      "Scope and System Boundaries",
+      "IT General Controls",
+      "Access Control and Segregation of Duties",
+      "Change Management",
+      "Audit Trail and Logging",
+      "Financial Reporting Controls"
+    ],
+    sectionGuidance: "Focus on Section 302 (CEO/CFO certification) and Section 404 (management assessment of internal controls). Identify which capabilities handle financial data, who can access them, and whether audit trails are complete."
+  },
+  ISO27001: {
+    fullName: "ISO/IEC 27001:2022 Information Security Management",
+    focusAreas: [
+      "Information security policies (Clause 5)",
+      "Risk assessment and treatment (Clause 6)",
+      "Information assets classification",
+      "Access control (Annex A 5.15\u20135.18)",
+      "Cryptography (Annex A 8.24)",
+      "Supplier relationships (Annex A 5.19\u20135.22)"
+    ],
+    requiredSections: [
+      "Scope and Context",
+      "Information Assets",
+      "Risk Assessment",
+      "Access Control",
+      "Cryptography and Data Protection",
+      "Incident Management",
+      "Business Continuity"
+    ],
+    sectionGuidance: "Structure the document around the ISO 27001:2022 Annex A controls. Map each capability to the relevant controls and identify statement of applicability (SoA) status."
+  },
+  GDPR: {
+    fullName: "General Data Protection Regulation (GDPR)",
+    focusAreas: [
+      "Article 30 Records of Processing Activities (ROPA)",
+      "Lawful basis for processing",
+      "Data subject rights (Articles 15\u201322)",
+      "Data retention and deletion",
+      "Data transfers (Chapter V)",
+      "Privacy by design and default (Article 25)"
+    ],
+    requiredSections: [
+      "Records of Processing Activities (ROPA)",
+      "Lawful Basis for Processing",
+      "Data Subject Rights",
+      "Data Retention Policy",
+      "International Data Transfers",
+      "Privacy by Design",
+      "Data Breach Response"
+    ],
+    sectionGuidance: "Structure around Article 30 ROPA format. For each capability processing personal data: identify purpose, lawful basis, categories of data, retention period, and third-party processors."
+  },
+  SOC2: {
+    fullName: "SOC 2 Type II (Trust Services Criteria)",
+    focusAreas: [
+      "Security (Common Criteria \u2014 CC)",
+      "Availability (Availability Criteria \u2014 A)",
+      "Confidentiality (Confidentiality Criteria \u2014 C)",
+      "Processing Integrity (PI)",
+      "Privacy (P)"
+    ],
+    requiredSections: [
+      "System Description",
+      "Security \u2014 Common Criteria",
+      "Availability",
+      "Confidentiality",
+      "Change Management",
+      "Monitoring",
+      "Vendor Management"
+    ],
+    sectionGuidance: "Follow the AICPA Trust Services Criteria. For each section, identify relevant capabilities, current controls, and any gaps versus the criteria."
+  },
+  CUSTOM: {
+    fullName: "Custom Compliance Framework",
+    focusAreas: [
+      "Scope and applicability",
+      "Control requirements",
+      "Risk assessment",
+      "Monitoring and audit"
+    ],
+    requiredSections: [
+      "Scope",
+      "Controls Inventory",
+      "Risk Assessment",
+      "Gaps and Recommendations"
+    ],
+    sectionGuidance: "Generate a structured compliance document based on the codebase capabilities and general best practices."
+  }
+};
+function buildCodebaseContext(blueprint, opts) {
+  const lines = [
+    `## Codebase Overview`,
+    `Project: ${opts.projectName}`,
+    `Total files: ${blueprint.stats.totalFiles}`,
+    `Dependency edges: ${blueprint.stats.runtimeEdges}`
+  ];
+  const topFiles = blueprint.files.slice().sort((a, b) => (b.importedByCount ?? 0) - (a.importedByCount ?? 0)).slice(0, 8);
+  lines.push(``, `## Most-imported files (likely critical components)`);
+  for (const f of topFiles) {
+    lines.push(`- ${f.path} [${f.category ?? "unknown"}] \u2014 imported by ${f.importedByCount ?? 0} modules`);
+  }
+  return lines.join("\n");
+}
+function buildAmberContext(amber) {
+  const lines = [`## AMBER Capability Registry`, amber.summary, ``];
+  for (const cap of amber.capabilities.slice(0, 20)) {
+    const tags = [cap.criticality, cap.lifecycle];
+    if (cap.driftCount > 0) tags.push(`drift:${cap.driftCount}`);
+    if (cap.hasTests) tags.push("documented");
+    lines.push(
+      `- ${cap.name} (${cap.id}) [${tags.join(", ")}]` + (cap.description ? `: ${cap.description}` : "")
+    );
+  }
+  return lines.join("\n");
+}
+function buildComplianceContext(ctx, framework) {
+  const lines = [
+    `## PRISM Compliance Analysis`,
+    `Framework: ${ctx.framework}`,
+    `Total capabilities: ${ctx.totalCapabilities}`,
+    `Risk summary \u2014 High: ${ctx.riskSummary.high}, Medium: ${ctx.riskSummary.medium}, Low: ${ctx.riskSummary.low}`,
+    ``,
+    `### Critical / High Capabilities`
+  ];
+  for (const cap of ctx.criticalCapabilities.slice(0, 15)) {
+    const tags = [cap.criticality];
+    if (cap.accessControl) tags.push("access-control");
+    if (cap.auditTrail) tags.push("audit-trail");
+    if (cap.dataProcessing) tags.push("data-processing");
+    if (!cap.testCoverage) tags.push("no-doc");
+    if (cap.driftCount > 0) tags.push(`drift:${cap.driftCount}`);
+    lines.push(`- ${cap.name} (${cap.id}) [${tags.join(", ")}]`);
+  }
+  if (framework === "SOX" && ctx.sox) {
+    lines.push(``, `### SOX: Financial capabilities: ${ctx.sox.financialCapabilities.join(", ") || "none identified"}`);
+    lines.push(`SOX: Access-controlled: ${ctx.sox.accessControlled.join(", ") || "none identified"}`);
+  }
+  if (framework === "GDPR" && ctx.gdpr) {
+    lines.push(``, `### GDPR: Data processing capabilities: ${ctx.gdpr.dataProcessingCapabilities.join(", ") || "none identified"}`);
+    lines.push(`GDPR: Retention capabilities: ${ctx.gdpr.retentionCapabilities.join(", ") || "none identified"}`);
+  }
+  if (framework === "ISO27001" && ctx.iso27001) {
+    lines.push(``, `### ISO27001: Information assets: ${ctx.iso27001.informationAssets.join(", ") || "none identified"}`);
+    lines.push(`ISO27001: Security controls: ${ctx.iso27001.securityControls.join(", ") || "none identified"}`);
+  }
+  if (framework === "SOC2" && ctx.soc2) {
+    lines.push(``, `### SOC2: Security capabilities: ${ctx.soc2.securityCapabilities.join(", ") || "none identified"}`);
+    lines.push(`SOC2: Availability capabilities: ${ctx.soc2.availabilityCapabilities.join(", ") || "none identified"}`);
+  }
+  return lines.join("\n");
+}
+function buildSystemPrompt8(opts) {
+  const toneMap = {
+    formal: "You are a senior compliance officer writing a formal regulatory compliance document for external auditors and regulators.",
+    technical: "You are a security-focused software architect writing a technical compliance assessment grounded in code-level evidence.",
+    executive: "You are a Chief Compliance Officer writing a concise executive summary compliance report for board-level review."
+  };
+  return `${toneMap[opts.tone]} Generate ONLY valid JSON \u2014 no markdown fences, no prose, no commentary outside the JSON object.`;
+}
+function buildUserPrompt8(opts) {
+  const spec = FRAMEWORK_SPECS[opts.framework] ?? FRAMEWORK_SPECS["CUSTOM"];
+  const frameworkName = opts.framework === "CUSTOM" ? opts.customFramework ?? "Custom Framework" : spec.fullName;
+  const contextParts = [
+    buildCodebaseContext(opts.blueprint, opts)
+  ];
+  if (opts.amberContext) contextParts.push(buildAmberContext(opts.amberContext));
+  if (opts.complianceContext) contextParts.push(buildComplianceContext(opts.complianceContext, opts.framework));
+  const org = opts.organizationName ? `Organization: ${opts.organizationName}` : "";
+  const recsInstruction = opts.includeRecommendations ? 'Include a "recommendations" array with 3\u20137 actionable remediation items.' : 'Set "recommendations" to an empty array.';
+  return [
+    `Generate a ${frameworkName} compliance document for: ${opts.projectName}`,
+    org,
+    `Tone: ${opts.tone}`,
+    ``,
+    contextParts.join("\n\n"),
+    ``,
+    `## Framework: ${frameworkName}`,
+    `Focus areas:`,
+    spec.focusAreas.map((f) => `- ${f}`).join("\n"),
+    ``,
+    spec.sectionGuidance,
+    ``,
+    `Required sections: ${spec.requiredSections.join(", ")}`,
+    ``,
+    `## Output Format`,
+    `Return a single JSON object with this exact shape:`,
+    `{`,
+    `  "documentTitle": "<formal document title>",`,
+    `  "executiveSummary": "<3\u20135 sentences summarising compliance posture>",`,
+    `  "confidentiality": "INTERNAL" | "CONFIDENTIAL" | "PUBLIC",`,
+    `  "complianceGaps": ["<gap 1>", "<gap 2>"],`,
+    `  "recommendations": ["<rec 1>", "<rec 2>"],`,
+    `  "sections": [`,
+    `    {`,
+    `      "title": "<section title>",`,
+    `      "content": "<substantive section prose, 2\u20134 paragraphs>",`,
+    `      "capabilities": ["<capability id 1>"],`,
+    `      "riskLevel": "high" | "medium" | "low",`,
+    `      "status": "compliant" | "partial" | "gap"`,
+    `    }`,
+    `  ]`,
+    `}`,
+    ``,
+    `Rules:`,
+    `- Derive ALL content from the codebase context above \u2014 no invented facts.`,
+    `- Each section MUST reference real capability IDs from the AMBER registry where available.`,
+    `- complianceGaps: list specific gaps found (missing controls, undocumented capabilities, drift).`,
+    recsInstruction,
+    `- confidentiality: CONFIDENTIAL for SOX/ISO27001/SOC2, INTERNAL for GDPR/CUSTOM.`,
+    `- Return ONLY valid JSON. No markdown. No prose outside the JSON.`
+  ].filter(Boolean).join("\n");
+}
+function buildFallback2(opts) {
+  const spec = FRAMEWORK_SPECS[opts.framework] ?? FRAMEWORK_SPECS["CUSTOM"];
+  const frameworkName = opts.framework === "CUSTOM" ? opts.customFramework ?? "Custom Framework" : spec.fullName;
+  const totalCaps = opts.complianceContext?.totalCapabilities ?? opts.amberContext?.capabilities.length ?? 0;
+  const highRisk = opts.complianceContext?.riskSummary.high ?? 0;
+  const sections = spec.requiredSections.map((title) => ({
+    title,
+    content: `Assessment of ${title.toLowerCase()} for ${opts.projectName}. Based on Blueprint analysis, ${totalCaps} capabilities were identified for review under ${frameworkName}.`,
+    capabilities: opts.amberContext?.capabilities.slice(0, 2).map((c) => c.id) ?? [],
+    riskLevel: "medium",
+    status: "partial"
+  }));
+  return {
+    framework: opts.framework,
+    documentTitle: `${opts.projectName} \u2014 ${frameworkName} Compliance Assessment`,
+    executiveSummary: `This document presents the ${frameworkName} compliance assessment for ${opts.projectName}. Analysis identified ${totalCaps} business capabilities, of which ${highRisk} present elevated risk requiring remediation. The assessment covers key control domains and provides a structured gap analysis.`,
+    sections,
+    complianceGaps: [
+      `${totalCaps - (opts.amberContext?.taggedFiles ?? 0)} files lack capability annotations \u2014 compliance traceability is incomplete.`,
+      `${opts.amberContext?.driftedCapabilities ?? 0} capabilities have documentation drift \u2014 controls may be misrepresented.`
+    ],
+    recommendations: opts.includeRecommendations ? [
+      "Ensure all critical capabilities have up-to-date @amber-doc annotations.",
+      "Implement automated compliance scanning in CI/CD pipeline.",
+      "Conduct quarterly capability review against framework requirements."
+    ] : [],
+    generatedAt: (/* @__PURE__ */ new Date()).toISOString(),
+    confidentiality: opts.framework === "GDPR" || opts.framework === "CUSTOM" ? "INTERNAL" : "CONFIDENTIAL"
+  };
+}
+function parseLlmResponse2(raw, opts) {
+  let text = raw.trim();
+  if (text.startsWith("```")) {
+    text = text.replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "").trim();
+  }
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  const rawSections = Array.isArray(parsed.sections) ? parsed.sections : [];
+  if (rawSections.length === 0) return null;
+  const sections = rawSections.map((s) => {
+    const sec = s ?? {};
+    return {
+      title: typeof sec.title === "string" ? sec.title : "Section",
+      content: typeof sec.content === "string" ? sec.content : "",
+      capabilities: Array.isArray(sec.capabilities) ? sec.capabilities.filter((c) => typeof c === "string") : [],
+      ...typeof sec.riskLevel === "string" && ["high", "medium", "low"].includes(sec.riskLevel) ? { riskLevel: sec.riskLevel } : {},
+      ...typeof sec.status === "string" && ["compliant", "partial", "gap"].includes(sec.status) ? { status: sec.status } : {}
+    };
+  });
+  const confidentiality = parsed.confidentiality === "CONFIDENTIAL" ? "CONFIDENTIAL" : parsed.confidentiality === "PUBLIC" ? "PUBLIC" : "INTERNAL";
+  return {
+    framework: opts.framework,
+    documentTitle: typeof parsed.documentTitle === "string" ? parsed.documentTitle : `${opts.projectName} \u2014 Compliance Assessment`,
+    sections,
+    executiveSummary: typeof parsed.executiveSummary === "string" ? parsed.executiveSummary : "",
+    complianceGaps: Array.isArray(parsed.complianceGaps) ? parsed.complianceGaps.filter((g) => typeof g === "string") : [],
+    recommendations: Array.isArray(parsed.recommendations) ? parsed.recommendations.filter((r) => typeof r === "string") : [],
+    generatedAt: (/* @__PURE__ */ new Date()).toISOString(),
+    confidentiality
+  };
+}
+async function generateComplianceDoc(opts) {
+  const systemPrompt = buildSystemPrompt8(opts);
+  const userPrompt = buildUserPrompt8(opts);
+  try {
+    const response = await opts.llm.complete({
+      systemPrompt,
+      messages: [{ role: "user", content: userPrompt }],
+      maxTokens: 6e3
+    });
+    const parsed = parseLlmResponse2(response.content, opts);
+    if (parsed) return parsed;
+  } catch {
+  }
+  return buildFallback2(opts);
+}
+
+// src/generators/generateADR.ts
+function buildCodebaseContext2(blueprint, projectName) {
+  const topFiles = blueprint.files.slice().sort((a, b) => (b.importedByCount ?? 0) - (a.importedByCount ?? 0)).slice(0, 8);
+  const totalLoc = blueprint.files.reduce((sum, f) => sum + (f.lineCount ?? 0), 0);
+  const lines = [
+    `## Codebase Overview`,
+    `Project: ${projectName}`,
+    `Total files: ${blueprint.stats.totalFiles}`,
+    `Total LOC: ${totalLoc.toLocaleString()}`,
+    `Dependency edges: ${blueprint.stats.runtimeEdges}`,
+    ``,
+    `## Most-imported files (architectural hotspots)`
+  ];
+  for (const f of topFiles) {
+    lines.push(`- ${f.path} [${f.category ?? "?"}] \u2014 imported by ${f.importedByCount ?? 0} modules, ${f.lineCount ?? 0} lines`);
+  }
+  if (blueprint.edges.length > 0) {
+    const edgeSample = blueprint.edges.slice(0, 8);
+    lines.push(``, `## Dependency sample`);
+    for (const e of edgeSample) {
+      lines.push(`- ${e.from} \u2192 ${e.to}${e.type ? ` (${e.type})` : ""}`);
+    }
+  }
+  return lines.join("\n");
+}
+function buildAdrContext(ctx, focus, selectedCycle, selectedCapability) {
+  const lines = [`## PRISM Analysis`];
+  if (ctx.coherenceScore !== null) {
+    lines.push(`Architecture Coherence Score: ${ctx.coherenceScore}/100`);
+  }
+  if (ctx.architecturalPatterns.length > 0) {
+    lines.push(`Detected patterns: ${ctx.architecturalPatterns.join(", ")}`);
+  }
+  if (ctx.topRisks.length > 0) {
+    lines.push(``, `### Top Architectural Risks`);
+    for (const r of ctx.topRisks.slice(0, 5)) {
+      lines.push(`- ${r.name} (${r.capabilityId}): ${r.reason}`);
+    }
+  }
+  if ((focus === "cycles" || focus === "overall") && ctx.importCycles.length > 0) {
+    lines.push(``, `### Import Cycles (${ctx.importCycles.length} detected)`);
+    const cyclesToShow = selectedCycle ? ctx.importCycles.filter((c) => c.files.some((f) => selectedCycle.includes(f))) : ctx.importCycles.slice(0, 5);
+    for (const c of cyclesToShow) {
+      lines.push(`- [${c.severity}] ${c.files.join(" \u2192 ")}`);
+    }
+  }
+  if ((focus === "capabilities" || focus === "overall") && ctx.capabilityBoundaries.length > 0) {
+    lines.push(``, `### Capability Boundaries (dependency edges)`);
+    const boundaries = selectedCapability ? ctx.capabilityBoundaries.filter((b) => b.from === selectedCapability || b.to === selectedCapability) : ctx.capabilityBoundaries.slice(0, 10);
+    for (const b of boundaries) {
+      lines.push(`- ${b.from} \u2192 ${b.to} (${b.edgeCount} import edges)`);
+    }
+  }
+  if ((focus === "dependencies" || focus === "overall") && ctx.highChurnFiles.length > 0) {
+    lines.push(``, `### High-Churn Files`);
+    for (const f of ctx.highChurnFiles.slice(0, 8)) {
+      const caps = f.capabilities.length > 0 ? ` [${f.capabilities.join(", ")}]` : "";
+      lines.push(`- ${f.path}${caps} \u2014 ${f.commits} commits`);
+    }
+  }
+  if (ctx.orphanedFiles.length > 0) {
+    lines.push(``, `### Orphaned Files (${ctx.orphanedFiles.length} untagged)`);
+    for (const f of ctx.orphanedFiles.slice(0, 5)) {
+      lines.push(`- ${f}`);
+    }
+  }
+  return lines.join("\n");
+}
+function buildAmberContext2(amber) {
+  const lines = [`## AMBER Capability Registry`, amber.summary, ``];
+  for (const cap of amber.capabilities.slice(0, 15)) {
+    const tags = [cap.criticality, cap.lifecycle];
+    if (cap.driftCount > 0) tags.push(`drift:${cap.driftCount}`);
+    lines.push(
+      `- ${cap.name} (${cap.id}) [${tags.join(", ")}]${cap.files.length > 0 ? ` \u2014 ${cap.files.length} files` : ""}`
+    );
+  }
+  return lines.join("\n");
+}
+function buildSystemPrompt9() {
+  return "You are a principal software architect with deep expertise in architecture decision records (ADRs) following the MADR (Markdown Architectural Decision Records) format. You analyse codebases and produce actionable, evidence-grounded ADRs. Generate ONLY valid JSON \u2014 no markdown fences, no prose, no commentary outside the JSON object.";
+}
+function buildUserPrompt9(opts) {
+  const focusDescriptions = {
+    overall: "Generate a comprehensive set of ADRs covering the most significant architectural decisions evident from import cycles, capability boundaries, dependency patterns, and code structure.",
+    cycles: "Focus on ADRs that address the import cycle patterns detected. Each cycle represents an implicit architectural decision \u2014 make those decisions explicit and propose resolutions.",
+    capabilities: "Focus on ADRs about capability boundary decisions \u2014 how capabilities are decomposed, what depends on what, and whether the boundaries are intentional or accidental.",
+    dependencies: "Focus on ADRs about dependency management \u2014 high-churn files, hotspots, fan-out patterns, and whether the dependency structure aligns with the intended architecture."
+  };
+  const parts = [
+    `Generate 3\u20137 Architecture Decision Records (ADRs) for: ${opts.projectName}`,
+    `Focus: ${opts.focus} \u2014 ${focusDescriptions[opts.focus]}`,
+    ``,
+    buildCodebaseContext2(opts.blueprint, opts.projectName)
+  ];
+  if (opts.amberContext) {
+    parts.push(``, buildAmberContext2(opts.amberContext));
+  }
+  if (opts.adrContext) {
+    parts.push(``, buildAdrContext(opts.adrContext, opts.focus, opts.selectedCycle, opts.selectedCapability));
+  }
+  if (opts.selectedCycle && opts.selectedCycle.length > 0) {
+    parts.push(``, `## Selected Cycle for Focus`, opts.selectedCycle.join(" \u2192 "));
+  }
+  if (opts.selectedCapability) {
+    parts.push(``, `## Selected Capability for Focus: ${opts.selectedCapability}`);
+  }
+  parts.push(
+    ``,
+    `## Output Format`,
+    `Return a single JSON object:`,
+    `{`,
+    `  "summary": "<2\u20133 sentence overall architectural assessment>",`,
+    `  "technicalDebt": "<1\u20132 sentence estimate of technical debt severity>",`,
+    `  "adrs": [`,
+    `    {`,
+    `      "id": "ADR-001",`,
+    `      "title": "<short imperative title>",`,
+    `      "date": "${(/* @__PURE__ */ new Date()).toISOString().slice(0, 10)}",`,
+    `      "status": "Proposed",`,
+    `      "context": "<what is the situation and why does this decision need to be made? 2\u20134 sentences>",`,
+    `      "decision": "<what was decided and why? 2\u20134 sentences with rationale>",`,
+    `      "consequences": {`,
+    `        "positive": ["<benefit 1>", "<benefit 2>"],`,
+    `        "negative": ["<drawback 1>"],`,
+    `        "neutral": ["<neutral fact>"]`,
+    `      },`,
+    `      "alternatives": ["<alternative option 1>", "<alternative option 2>"],`,
+    `      "relatedCapabilities": ["<capability-id-1>"],`,
+    `      "relatedFiles": ["<file/path.ts>"]`,
+    `    }`,
+    `  ]`,
+    `}`,
+    ``,
+    `Rules:`,
+    `- IDs: ADR-001, ADR-002, \u2026 in order.`,
+    `- Derive ALL content from the codebase data above \u2014 no invented facts.`,
+    `- Each ADR MUST address a real pattern visible in the data (cycle, boundary, dependency).`,
+    `- relatedCapabilities: reference real capability IDs from AMBER registry when available.`,
+    `- relatedFiles: reference real file paths from the blueprint.`,
+    `- status is always "Proposed" for new ADRs.`,
+    `- Return ONLY valid JSON. No markdown. No prose outside the JSON.`
+  );
+  return parts.join("\n");
+}
+function buildFallback3(opts) {
+  const today = (/* @__PURE__ */ new Date()).toISOString().slice(0, 10);
+  const adrs = [];
+  if (opts.focus === "cycles" || opts.focus === "overall") {
+    const cycleFiles = opts.adrContext?.importCycles[0]?.files ?? opts.selectedCycle ?? [];
+    adrs.push({
+      id: "ADR-001",
+      title: "Introduce Module Boundary Abstractions to Break Import Cycles",
+      date: today,
+      status: "Proposed",
+      context: `${opts.projectName} contains ${opts.adrContext?.importCycles.length ?? "multiple"} detected import cycles. Circular dependencies create tight coupling that makes independent testing and deployment difficult.`,
+      decision: "Introduce shared interface/abstraction modules at cycle breakpoints to invert dependencies. Files that create cycles should depend on abstractions, not on each other directly.",
+      consequences: {
+        positive: ["Enables independent testing of cyclic modules", "Reduces coupling between layers"],
+        negative: ["Requires refactoring existing imports", "Short-term increase in file count"],
+        neutral: ["Aligns with Dependency Inversion Principle"]
+      },
+      alternatives: ["Extract shared state to a separate module", "Use event-driven communication between modules"],
+      relatedCapabilities: opts.amberContext?.capabilities.slice(0, 2).map((c) => c.id) ?? [],
+      relatedFiles: cycleFiles.slice(0, 3)
+    });
+  }
+  if (opts.focus === "capabilities" || opts.focus === "overall") {
+    adrs.push({
+      id: `ADR-00${adrs.length + 1}`,
+      title: "Enforce Explicit Capability Boundaries with Barrel Exports",
+      date: today,
+      status: "Proposed",
+      context: `Capability boundaries in ${opts.projectName} are implicit \u2014 cross-capability imports create unintended dependencies. The AMBER registry identifies capabilities but import boundaries don't enforce them.`,
+      decision: "Each capability should expose a single barrel index file. External modules import only from the barrel, never from internal implementation files.",
+      consequences: {
+        positive: ["Capability-level API surfaces become explicit", "Refactoring internals doesn't break consumers"],
+        negative: ["Requires creating barrel files for each capability", "Barrel files need maintenance"],
+        neutral: ["Standard pattern in large TypeScript codebases"]
+      },
+      alternatives: ["Use ESLint import boundaries rules", "Monorepo packages per capability"],
+      relatedCapabilities: opts.amberContext?.capabilities.slice(0, 3).map((c) => c.id) ?? [],
+      relatedFiles: []
+    });
+  }
+  adrs.push({
+    id: `ADR-00${adrs.length + 1}`,
+    title: "Establish Architectural Fitness Functions for Ongoing Compliance",
+    date: today,
+    status: "Proposed",
+    context: `The ${opts.projectName} codebase currently lacks automated enforcement of architectural constraints. Without fitness functions, architectural drift accumulates silently.`,
+    decision: "Integrate PRISM / prism0x2A scans into CI to track coherence score, import cycles, and capability drift. Gate merges on cycle count not increasing.",
+    consequences: {
+      positive: ["Architectural health is continuously measured", "Teams get early warning of degradation"],
+      negative: ["CI time increases slightly", "Requires PRISM setup in CI environment"],
+      neutral: ["Shift from reactive to proactive architecture governance"]
+    },
+    alternatives: ["Manual architecture review cadence", "Weekly automated reports only"],
+    relatedCapabilities: [],
+    relatedFiles: []
+  });
+  return {
+    adrs,
+    summary: `${opts.projectName} has ${opts.blueprint.stats.totalFiles} files with ${opts.blueprint.stats.runtimeEdges} dependency edges. ${opts.adrContext?.importCycles.length ?? 0} import cycles and ${opts.adrContext?.capabilityBoundaries.length ?? 0} cross-capability dependency edges indicate structural decisions that should be made explicit via ADRs.`,
+    technicalDebt: "Moderate-to-high architectural debt evident from cycle patterns and implicit capability boundaries. Estimated 3\u20136 engineering weeks to implement the proposed decisions."
+  };
+}
+function parseLlmResponse3(raw) {
+  let text = raw.trim();
+  if (text.startsWith("```")) {
+    text = text.replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "").trim();
+  }
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  const rawAdrs = Array.isArray(parsed.adrs) ? parsed.adrs : [];
+  if (rawAdrs.length === 0) return null;
+  const today = (/* @__PURE__ */ new Date()).toISOString().slice(0, 10);
+  const adrs = rawAdrs.map((a, idx) => {
+    const adr = a ?? {};
+    const consequences = typeof adr.consequences === "object" && adr.consequences !== null ? adr.consequences : {};
+    const status = typeof adr.status === "string" && ["Proposed", "Accepted", "Deprecated", "Superseded"].includes(adr.status) ? adr.status : "Proposed";
+    return {
+      id: typeof adr.id === "string" ? adr.id : `ADR-${String(idx + 1).padStart(3, "0")}`,
+      title: typeof adr.title === "string" ? adr.title : `Decision ${idx + 1}`,
+      date: typeof adr.date === "string" ? adr.date : today,
+      status,
+      context: typeof adr.context === "string" ? adr.context : "",
+      decision: typeof adr.decision === "string" ? adr.decision : "",
+      consequences: {
+        positive: Array.isArray(consequences.positive) ? consequences.positive.filter((s) => typeof s === "string") : [],
+        negative: Array.isArray(consequences.negative) ? consequences.negative.filter((s) => typeof s === "string") : [],
+        neutral: Array.isArray(consequences.neutral) ? consequences.neutral.filter((s) => typeof s === "string") : []
+      },
+      alternatives: Array.isArray(adr.alternatives) ? adr.alternatives.filter((s) => typeof s === "string") : [],
+      relatedCapabilities: Array.isArray(adr.relatedCapabilities) ? adr.relatedCapabilities.filter((s) => typeof s === "string") : [],
+      relatedFiles: Array.isArray(adr.relatedFiles) ? adr.relatedFiles.filter((s) => typeof s === "string") : []
+    };
+  });
+  return {
+    adrs,
+    summary: typeof parsed.summary === "string" ? parsed.summary : "",
+    technicalDebt: typeof parsed.technicalDebt === "string" ? parsed.technicalDebt : ""
+  };
+}
+async function generateADR(opts) {
+  try {
+    const response = await opts.llm.complete({
+      systemPrompt: buildSystemPrompt9(),
+      messages: [{ role: "user", content: buildUserPrompt9(opts) }],
+      maxTokens: 6e3
+    });
+    const parsed = parseLlmResponse3(response.content);
+    if (parsed) return parsed;
+  } catch {
+  }
+  return buildFallback3(opts);
+}
+
+// src/generators/generateSprintRetro.ts
+function classifyCommit(subject) {
+  const s = subject.toLowerCase();
+  if (s.startsWith("feat") || s.includes("add ") || s.includes("implement") || s.includes("new ")) return "feature";
+  if (s.startsWith("fix") || s.includes("bug") || s.includes("patch") || s.includes("hotfix")) return "fix";
+  if (s.startsWith("refactor") || s.startsWith("refact") || s.includes("rewrite") || s.includes("restructure")) return "refactor";
+  return "chore";
+}
+function buildContext2(opts) {
+  const lines = [
+    `## Sprint Context`,
+    `Project: ${opts.projectName}`,
+    `Sprint: ${opts.sprintName ?? "Current Sprint"}`,
+    `Team size: ${opts.teamSize ?? "unknown"}`,
+    `Tone: ${opts.tone}`,
+    ``
+  ];
+  if (opts.gitContext) {
+    const { commits, filesChanged, fromRef, toRef } = opts.gitContext;
+    lines.push(
+      `## Git Activity (${fromRef} \u2192 ${toRef})`,
+      `Commits: ${commits.length}`,
+      `Files changed: ${filesChanged}`,
+      ``,
+      `### Commits`,
+      ...commits.slice(0, 30).map(
+        (c) => `- [${classifyCommit(c.subject)}] ${c.subject}${c.author ? ` \u2014 ${c.author}` : ""}`
+      ),
+      ``
+    );
+  }
+  if (opts.scoreBefore !== void 0 || opts.scoreAfter !== void 0) {
+    lines.push(
+      `## Architecture Health`,
+      `Score at sprint start: ${opts.scoreBefore ?? "unknown"}`,
+      `Score at sprint end: ${opts.scoreAfter ?? "unknown"}`,
+      ``
+    );
+  }
+  if (opts.amberContext) {
+    const { capabilities, driftedCapabilities, summary } = opts.amberContext;
+    lines.push(
+      `## AMBER Capability Layer`,
+      summary,
+      `Drifted capabilities: ${driftedCapabilities}`,
+      ``,
+      `### Capabilities`,
+      ...capabilities.slice(0, 15).map(
+        (c) => `- ${c.name} [${c.criticality}/${c.lifecycle}]${c.driftCount > 0 ? ` \u26A0 drift: ${c.driftCount} files` : ""}`
+      ),
+      ``
+    );
+  }
+  if (opts.greenContext) {
+    lines.push(
+      `## GREEN Cross-Layer Insights`,
+      opts.greenContext.summary,
+      ``
+    );
+    if (opts.greenContext.topRisks.length > 0) {
+      lines.push(
+        `### Top Risks`,
+        ...opts.greenContext.topRisks.slice(0, 5).map(
+          (r) => `- ${r.name}: ${r.reason} (impact: ${r.impact}/10)`
+        ),
+        ``
+      );
+    }
+  }
+  return lines.join("\n");
+}
+function buildSystemPrompt10(opts) {
+  const toneMap = {
+    professional: "You are a senior engineering manager facilitating a data-driven sprint retrospective. Be precise and actionable.",
+    casual: "You are a friendly scrum master running a fun and honest retrospective. Keep it light but insightful.",
+    "team-friendly": "You are a team lead who wants everyone to feel heard. Balance honesty with encouragement."
+  };
+  return `${toneMap[opts.tone]} Generate ONLY valid JSON \u2014 no markdown fences, no prose, no commentary outside the JSON object.`;
+}
+function buildUserPrompt10(opts) {
+  const context = buildContext2(opts);
+  const sprintName = opts.sprintName ?? "Current Sprint";
+  return [
+    `Generate a sprint retrospective for: ${opts.projectName} \u2014 ${sprintName}`,
+    ``,
+    context,
+    ``,
+    `## Output Format`,
+    `Return a single JSON object with this exact shape:`,
+    `{`,
+    `  "sprintName": "${sprintName}",`,
+    `  "period": "<date range or sprint identifier>",`,
+    `  "summary": "<2-3 sentence executive summary of the sprint>",`,
+    `  "delivered": [`,
+    `    { "item": "<what was built>", "capability": "<AMBER capability id if applicable>", "type": "feature|fix|refactor|chore" }`,
+    `  ],`,
+    `  "healthDelta": {`,
+    `    "scoreBefore": <number or null>,`,
+    `    "scoreAfter": <number or null>,`,
+    `    "delta": <number or null>,`,
+    `    "verdict": "improved|stable|degraded",`,
+    `    "degradedCapabilities": ["<capability name>"]`,
+    `  },`,
+    `  "wentWell": ["<positive thing 1>", ...],`,
+    `  "improvements": ["<improvement needed 1>", ...],`,
+    `  "puzzles": ["<question or confusion 1>", ...],`,
+    `  "actions": ["<concrete action item 1>", ...],`,
+    `  "debtIncurred": ["<technical debt item 1>", ...],`,
+    `  "slideTitle": "<1 sentence sprint summary for a slide>",`,
+    `  "slidePoints": ["<key point 1>", "<key point 2>", "<key point 3>"]`,
+    `}`,
+    ``,
+    `Rules:`,
+    `- delivered: derive from git commits, 5\u201315 items`,
+    `- wentWell: 3\u20135 genuine positives from the data`,
+    `- improvements: 3\u20135 concrete improvements (not vague)`,
+    `- puzzles: 2\u20134 things that were unclear or confusing this sprint`,
+    `- actions: 3\u20135 specific, assigned-sounding action items`,
+    `- debtIncurred: 1\u20134 technical debt items visible from the commits/drift`,
+    `- slidePoints: exactly 3 bullet strings \u2264 80 characters`,
+    `- Derive everything from the context above \u2014 no invented facts`,
+    `- Return ONLY valid JSON. No markdown. No prose outside the JSON.`
+  ].join("\n");
+}
+function computeHealthDelta(scoreBefore, scoreAfter, amberContext) {
+  const before = scoreBefore ?? null;
+  const after = scoreAfter ?? null;
+  const delta = before !== null && after !== null ? after - before : null;
+  let verdict = "stable";
+  if (delta !== null) {
+    if (delta > 2) verdict = "improved";
+    else if (delta < -2) verdict = "degraded";
+  }
+  const degradedCapabilities = amberContext ? amberContext.capabilities.filter((c) => c.driftCount > 0 && (c.criticality === "critical" || c.criticality === "high")).map((c) => c.name) : [];
+  return { scoreBefore: before, scoreAfter: after, delta, verdict, degradedCapabilities };
+}
+function buildFallback4(opts) {
+  const sprintName = opts.sprintName ?? "Current Sprint";
+  const commits = opts.gitContext?.commits ?? [];
+  const healthDelta = computeHealthDelta(opts.scoreBefore, opts.scoreAfter, opts.amberContext);
+  const delivered = commits.slice(0, 10).map((c) => ({
+    item: c.subject,
+    type: classifyCommit(c.subject)
+  }));
+  return {
+    sprintName,
+    period: opts.gitContext ? `${opts.gitContext.fromRef} \u2192 ${opts.gitContext.toRef}` : "This sprint",
+    summary: `${opts.projectName} completed ${commits.length} commits this sprint across ${opts.gitContext?.filesChanged ?? 0} files.`,
+    delivered,
+    healthDelta,
+    wentWell: [
+      `${commits.filter((c) => classifyCommit(c.subject) === "feature").length} features shipped`,
+      `${commits.filter((c) => classifyCommit(c.subject) === "fix").length} bugs fixed`,
+      "Team maintained delivery cadence"
+    ],
+    improvements: [
+      "Reduce WIP by completing in-flight work before starting new",
+      "Improve commit message clarity",
+      "Address documentation drift in critical capabilities"
+    ],
+    puzzles: [
+      "Unclear ownership for cross-capability changes",
+      "Test coverage strategy for new features"
+    ],
+    actions: [
+      "Schedule architecture review for flagged capabilities",
+      "Set up automated coherence score tracking in CI",
+      "Document top 3 architectural decisions made this sprint"
+    ],
+    debtIncurred: healthDelta.degradedCapabilities.length > 0 ? healthDelta.degradedCapabilities.map((c) => `Documentation drift in ${c}`) : ["Review and address any undocumented capability changes"],
+    slideTitle: `${sprintName}: ${commits.length} commits, ${opts.gitContext?.filesChanged ?? 0} files changed`,
+    slidePoints: [
+      `${commits.length} commits across ${opts.gitContext?.filesChanged ?? 0} files`,
+      `Architecture health: ${healthDelta.verdict}`,
+      `${healthDelta.degradedCapabilities.length} capabilities need attention`
+    ]
+  };
+}
+function parseLlmResponse4(raw, opts) {
+  let text = raw.trim();
+  if (text.startsWith("```")) {
+    text = text.replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "").trim();
+  }
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  const sprintName = typeof parsed.sprintName === "string" ? parsed.sprintName : opts.sprintName ?? "Current Sprint";
+  const period = typeof parsed.period === "string" ? parsed.period : "";
+  const summary = typeof parsed.summary === "string" ? parsed.summary : "";
+  const rawDelivered = Array.isArray(parsed.delivered) ? parsed.delivered : [];
+  const delivered = rawDelivered.filter((d) => typeof d.item === "string").map((d) => {
+    const item = d;
+    const type = ["feature", "fix", "refactor", "chore"].includes(item.type) ? item.type : "chore";
+    return {
+      item: item.item,
+      ...typeof item.capability === "string" ? { capability: item.capability } : {},
+      type
+    };
+  });
+  const hd = parsed.healthDelta ?? {};
+  const scoreBeforeRaw = typeof hd.scoreBefore === "number" ? hd.scoreBefore : null;
+  const scoreAfterRaw = typeof hd.scoreAfter === "number" ? hd.scoreAfter : null;
+  const deltaRaw = typeof hd.delta === "number" ? hd.delta : null;
+  const verdictRaw = ["improved", "stable", "degraded"].includes(hd.verdict) ? hd.verdict : "stable";
+  const degradedCaps = Array.isArray(hd.degradedCapabilities) ? hd.degradedCapabilities.filter((x) => typeof x === "string") : [];
+  const toStringArray = (key) => {
+    const val = parsed[key];
+    if (!Array.isArray(val)) return [];
+    return val.filter((x) => typeof x === "string");
+  };
+  if (!summary && delivered.length === 0) return null;
+  return {
+    sprintName,
+    period,
+    summary,
+    delivered,
+    healthDelta: {
+      scoreBefore: scoreBeforeRaw,
+      scoreAfter: scoreAfterRaw,
+      delta: deltaRaw,
+      verdict: verdictRaw,
+      degradedCapabilities: degradedCaps
+    },
+    wentWell: toStringArray("wentWell"),
+    improvements: toStringArray("improvements"),
+    puzzles: toStringArray("puzzles"),
+    actions: toStringArray("actions"),
+    debtIncurred: toStringArray("debtIncurred"),
+    slideTitle: typeof parsed.slideTitle === "string" ? parsed.slideTitle : sprintName,
+    slidePoints: toStringArray("slidePoints").slice(0, 3)
+  };
+}
+async function generateSprintRetro(opts) {
+  try {
+    const response = await opts.llm.complete({
+      systemPrompt: buildSystemPrompt10(opts),
+      messages: [{ role: "user", content: buildUserPrompt10(opts) }],
+      maxTokens: 4096
+    });
+    const parsed = parseLlmResponse4(response.content, opts);
+    if (parsed) return parsed;
+  } catch {
+  }
+  return buildFallback4(opts);
+}
+
+// src/generators/generateNewsletter.ts
+function buildContext3(opts) {
+  const { digestContext: d, amberContext, projectName, teamName, targetAudience, tone } = opts;
+  const lines = [
+    `## Newsletter Context`,
+    `Project: ${projectName}`,
+    `Team: ${teamName ?? "Engineering"}`,
+    `Target audience: ${targetAudience}`,
+    `Tone: ${tone}`,
+    ``,
+    `## Weekly Digest \u2014 ${d.period}`,
+    `Commits: ${d.commitCount}`,
+    `Files changed: ${d.filesChanged}`,
+    `Coherence grade: ${d.grade ?? "N/A"}`,
+    `Score start: ${d.scoreStart ?? "N/A"}`,
+    `Score end: ${d.scoreEnd ?? "N/A"}`,
+    `Score delta: ${d.scoreDelta !== null ? d.scoreDelta > 0 ? `+${d.scoreDelta}` : String(d.scoreDelta) : "N/A"}`,
+    `Health summary: ${d.healthSummary}`,
+    ``
+  ];
+  if (d.topCommits.length > 0) {
+    lines.push(`### Top commits`, ...d.topCommits.map((c) => `- ${c}`), ``);
+  }
+  if (d.newDrifts.length > 0) {
+    lines.push(`### New capability drifts`, ...d.newDrifts.map((c) => `- ${c}`), ``);
+  }
+  if (d.resolvedDrifts.length > 0) {
+    lines.push(`### Resolved drifts`, ...d.resolvedDrifts.map((c) => `- ${c}`), ``);
+  }
+  if (d.newRisks.length > 0) {
+    lines.push(`### New risks`, ...d.newRisks.map((r) => `- ${r}`), ``);
+  }
+  if (amberContext) {
+    lines.push(`## AMBER Capability Summary`, amberContext.summary, ``);
+  }
+  return lines.join("\n");
+}
+function buildSystemPrompt11(opts) {
+  const audienceMap = {
+    "product-team": "You are writing a newsletter for the product engineering team \u2014 technical details are welcome.",
+    management: "You are writing a newsletter for engineering managers \u2014 focus on velocity, quality, and risk.",
+    stakeholders: "You are writing a newsletter for business stakeholders \u2014 translate technical work into business outcomes.",
+    "all-hands": "You are writing a company-wide newsletter \u2014 accessible to everyone, inspiring, and concise."
+  };
+  const toneMap = {
+    professional: "Use a professional, data-driven tone.",
+    casual: "Use a friendly, conversational tone \u2014 contractions welcome.",
+    accessible: "Use zero technical jargon. If you must mention a technical concept, define it in plain language immediately."
+  };
+  return [
+    audienceMap[opts.targetAudience],
+    toneMap[opts.tone],
+    "Generate ONLY valid JSON \u2014 no markdown fences, no prose, no commentary outside the JSON object."
+  ].join(" ");
+}
+function buildUserPrompt11(opts) {
+  const context = buildContext3(opts);
+  const { projectName, teamName, targetAudience, tone, includeMetrics } = opts;
+  return [
+    `Generate a weekly engineering newsletter for: ${projectName}`,
+    `Team: ${teamName ?? "Engineering"} | Audience: ${targetAudience} | Tone: ${tone}`,
+    `Include metrics: ${includeMetrics}`,
+    ``,
+    context,
+    ``,
+    `## Output Format`,
+    `Return a single JSON object with this exact shape:`,
+    `{`,
+    `  "subject": "<email subject line, \u226460 chars>",`,
+    `  "preview": "<email preview text, \u226450 chars>",`,
+    `  "greeting": "<opening greeting>",`,
+    `  "headline": "<1-sentence hook that opens the newsletter>",`,
+    `  "sections": [`,
+    `    { "heading": "<section heading>", "body": "<2-4 sentences>", "type": "highlight|metrics|risk|shoutout|upcoming" }`,
+    `  ],`,
+    `  "metrics": [`,
+    `    { "label": "<metric name>", "value": "<value>", "trend": "up|down|stable" }`,
+    `  ],`,
+    `  "closing": "<1-2 sentence closing>",`,
+    `  "unsubscribeNote": "Reply STOP to unsubscribe from this digest.",`,
+    `  "slackVersion": "<Slack mrkdwn formatted version of the newsletter>",`,
+    `  "teamsVersion": "<Microsoft Teams Adaptive Card JSON as a string>",`,
+    `  "htmlVersion": "<complete HTML email with inline styles, dark-mode friendly>"`,
+    `}`,
+    ``,
+    `Rules:`,
+    `- sections: 3\u20135 sections covering the highlights`,
+    `- metrics: ${includeMetrics ? "include 3\u20135 key metrics from the digest" : "return empty array []"}`,
+    `- slackVersion: use *bold*, _italic_, \`code\`, > blockquote, \u2022 bullets (mrkdwn)`,
+    `- teamsVersion: valid Teams Adaptive Card JSON with TextBlock, FactSet elements`,
+    `- htmlVersion: full HTML with <html><head><body>, inline CSS, no external resources`,
+    `- subject: must reference the week/project and be compelling`,
+    `- Derive all content from the digest context above \u2014 no invented facts`,
+    `- Return ONLY valid JSON. No markdown fences. No prose outside the JSON.`
+  ].join("\n");
+}
+function buildHtmlEmail(opts, sections, metrics) {
+  const { projectName, teamName, digestContext: d } = opts;
+  const sectionHtml = sections.map((s) => `
+    <tr><td style="padding:20px 32px 0">
+      <h2 style="margin:0 0 8px;font-size:16px;color:#e2e8f0">${s.heading}</h2>
+      <p style="margin:0;color:#94a3b8;font-size:14px;line-height:1.6">${s.body}</p>
+    </td></tr>`).join("");
+  const metricsHtml = metrics.length > 0 ? `
+    <tr><td style="padding:20px 32px 0">
+      <table width="100%" cellpadding="0" cellspacing="0" border="0">
+        <tr>${metrics.map((m) => `
+          <td style="text-align:center;padding:12px;background:#1e293b;border-radius:8px;margin:4px">
+            <div style="font-size:22px;font-weight:700;color:${m.trend === "up" ? "#34d399" : m.trend === "down" ? "#f87171" : "#94a3b8"}">${m.value}</div>
+            <div style="font-size:12px;color:#64748b;margin-top:4px">${m.label}</div>
+          </td>`).join("<td style='width:8px'></td>")}</tr>
+      </table>
+    </td></tr>` : "";
+  return `<!DOCTYPE html>
+<html lang="en">
+<head><meta charset="UTF-8"><meta name="viewport" content="width=device-width,initial-scale=1.0"><title>${projectName} \u2014 Weekly Digest</title></head>
+<body style="margin:0;padding:0;background:#0f172a;font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',sans-serif">
+  <table width="100%" cellpadding="0" cellspacing="0" border="0" style="max-width:600px;margin:0 auto">
+    <tr><td style="padding:32px 32px 16px;border-bottom:1px solid #1e293b">
+      <h1 style="margin:0;font-size:24px;font-weight:700;color:#f1f5f9">${projectName}</h1>
+      <p style="margin:4px 0 0;color:#64748b;font-size:14px">${teamName ?? "Engineering"} \xB7 ${d.period}</p>
+    </td></tr>
+    ${sectionHtml}
+    ${metricsHtml}
+    <tr><td style="padding:24px 32px;border-top:1px solid #1e293b;margin-top:24px">
+      <p style="margin:0;color:#475569;font-size:12px">Generated by forge0x2B \xB7 Reply STOP to unsubscribe.</p>
+    </td></tr>
+  </table>
+</body>
+</html>`;
+}
+function buildSlackVersion(opts, sections, metrics) {
+  const { projectName, digestContext: d } = opts;
+  const lines = [
+    `*${projectName} \u2014 Weekly Engineering Digest*`,
+    `_${d.period}_`,
+    ``,
+    `*Health:* ${d.healthSummary}`,
+    ``,
+    ...sections.map((s) => [`*${s.heading}*`, s.body, ""].join("\n"))
+  ];
+  if (metrics.length > 0) {
+    lines.push(
+      `*Key Metrics*`,
+      ...metrics.map((m) => `\u2022 *${m.label}:* ${m.value} ${m.trend === "up" ? "\u2191" : m.trend === "down" ? "\u2193" : "\u2192"}`),
+      ``
+    );
+  }
+  lines.push(`_Generated by forge0x2B_`);
+  return lines.join("\n");
+}
+function buildTeamsVersion(opts, sections) {
+  const { projectName, digestContext: d } = opts;
+  const card = {
+    type: "AdaptiveCard",
+    $schema: "http://adaptivecards.io/schemas/adaptive-card.json",
+    version: "1.4",
+    body: [
+      {
+        type: "TextBlock",
+        text: `${projectName} \u2014 Weekly Engineering Digest`,
+        weight: "Bolder",
+        size: "Large",
+        color: "Accent"
+      },
+      {
+        type: "TextBlock",
+        text: d.period,
+        isSubtle: true,
+        spacing: "None"
+      },
+      {
+        type: "TextBlock",
+        text: d.healthSummary,
+        wrap: true,
+        spacing: "Medium"
+      },
+      ...sections.slice(0, 3).map((s) => ({
+        type: "Container",
+        spacing: "Medium",
+        items: [
+          { type: "TextBlock", text: s.heading, weight: "Bolder", wrap: true },
+          { type: "TextBlock", text: s.body, wrap: true, isSubtle: true }
+        ]
+      })),
+      {
+        type: "TextBlock",
+        text: "Generated by forge0x2B",
+        isSubtle: true,
+        size: "Small",
+        spacing: "Large"
+      }
+    ]
+  };
+  return JSON.stringify(card);
+}
+function buildFallback5(opts) {
+  const { projectName, teamName, digestContext: d, includeMetrics } = opts;
+  const sections = [
+    {
+      heading: "This Week in Engineering",
+      body: `The team shipped ${d.commitCount} commits across ${d.filesChanged} files during ${d.period}. ${d.healthSummary}`,
+      type: "highlight"
+    }
+  ];
+  if (d.topCommits.length > 0) {
+    sections.push({
+      heading: "Highlights",
+      body: d.topCommits.slice(0, 3).join(". ") + ".",
+      type: "highlight"
+    });
+  }
+  if (d.newRisks.length > 0) {
+    sections.push({
+      heading: "On Our Radar",
+      body: `We identified ${d.newRisks.length} new risk${d.newRisks.length > 1 ? "s" : ""} this week: ${d.newRisks.slice(0, 2).join(", ")}.`,
+      type: "risk"
+    });
+  }
+  const metrics = includeMetrics ? [
+    { label: "Commits", value: String(d.commitCount), trend: "stable" },
+    { label: "Files Changed", value: String(d.filesChanged), trend: "stable" },
+    ...d.grade ? [{ label: "Architecture Grade", value: d.grade, trend: d.scoreDelta !== null && d.scoreDelta > 0 ? "up" : d.scoreDelta !== null && d.scoreDelta < 0 ? "down" : "stable" }] : []
+  ] : [];
+  return {
+    subject: `${projectName} Engineering Digest \u2014 ${d.period}`,
+    preview: `${d.commitCount} commits \xB7 ${d.healthSummary.slice(0, 40)}`,
+    greeting: `Hello ${teamName ?? "team"},`,
+    headline: `Here's what the engineering team shipped this week.`,
+    sections,
+    metrics,
+    closing: `Thanks for reading. See you next week!`,
+    unsubscribeNote: "Reply STOP to unsubscribe from this digest.",
+    slackVersion: buildSlackVersion(opts, sections, metrics),
+    teamsVersion: buildTeamsVersion(opts, sections),
+    htmlVersion: buildHtmlEmail(opts, sections, metrics)
+  };
+}
+function parseLlmResponse5(raw, opts) {
+  let text = raw.trim();
+  if (text.startsWith("```")) {
+    text = text.replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "").trim();
+  }
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  const getString = (key, fallback) => typeof parsed[key] === "string" ? parsed[key] : fallback;
+  const rawSections = Array.isArray(parsed.sections) ? parsed.sections : [];
+  const sections = rawSections.filter((s) => typeof s.heading === "string").map((s) => {
+    const item = s;
+    return {
+      heading: item.heading,
+      body: typeof item.body === "string" ? item.body : "",
+      type: ["highlight", "metrics", "risk", "shoutout", "upcoming"].includes(item.type) ? item.type : "highlight"
+    };
+  });
+  const rawMetrics = Array.isArray(parsed.metrics) ? parsed.metrics : [];
+  const metrics = rawMetrics.filter((m) => typeof m.label === "string").map((m) => {
+    const item = m;
+    return {
+      label: item.label,
+      value: typeof item.value === "string" ? item.value : "",
+      trend: ["up", "down", "stable"].includes(item.trend) ? item.trend : "stable"
+    };
+  });
+  const subject = getString("subject", "");
+  if (!subject) return null;
+  const slackVersion = getString("slackVersion", "") || buildSlackVersion(opts, sections, metrics);
+  const teamsVersion = getString("teamsVersion", "") || buildTeamsVersion(opts, sections);
+  const htmlVersion = getString("htmlVersion", "") || buildHtmlEmail(opts, sections, metrics);
+  return {
+    subject,
+    preview: getString("preview", "").slice(0, 50),
+    greeting: getString("greeting", `Hello ${opts.teamName ?? "team"},`),
+    headline: getString("headline", ""),
+    sections,
+    metrics,
+    closing: getString("closing", ""),
+    unsubscribeNote: "Reply STOP to unsubscribe from this digest.",
+    slackVersion,
+    teamsVersion,
+    htmlVersion
+  };
+}
+async function generateNewsletter(opts) {
+  try {
+    const response = await opts.llm.complete({
+      systemPrompt: buildSystemPrompt11(opts),
+      messages: [{ role: "user", content: buildUserPrompt11(opts) }],
+      maxTokens: 8192
+    });
+    const parsed = parseLlmResponse5(response.content, opts);
+    if (parsed) return parsed;
+  } catch {
+  }
+  return buildFallback5(opts);
+}
+
+// src/generators/generateRadio.ts
+function buildContext4(opts) {
+  const { digestContext: d, amberContext, projectName, audience } = opts;
+  const lines = [
+    `## Architecture Radio Context`,
+    `Project: ${projectName}`,
+    `Audience: ${audience}`,
+    `Period: ${d.period}`,
+    ``,
+    `## Weekly Digest`,
+    `Commits: ${d.commitCount}`,
+    `Files changed: ${d.filesChanged}`,
+    `Coherence grade: ${d.grade ?? "N/A"}`,
+    `Score: ${d.scoreStart ?? "N/A"} \u2192 ${d.scoreEnd ?? "N/A"} (delta: ${d.scoreDelta !== null ? d.scoreDelta > 0 ? `+${d.scoreDelta}` : String(d.scoreDelta) : "N/A"})`,
+    `Health: ${d.healthSummary}`,
+    ``
+  ];
+  if (d.topCommits.length > 0) {
+    lines.push(`### Top Commits`, ...d.topCommits.slice(0, 5).map((c) => `- ${c}`), ``);
+  }
+  if (d.newDrifts.length > 0) {
+    lines.push(`### Capability Drifts`, ...d.newDrifts.map((c) => `- ${c}`), ``);
+  }
+  if (d.resolvedDrifts.length > 0) {
+    lines.push(`### Resolved Drifts`, ...d.resolvedDrifts.map((c) => `- ${c}`), ``);
+  }
+  if (d.newRisks.length > 0) {
+    lines.push(`### New Risks`, ...d.newRisks.map((r) => `- ${r}`), ``);
+  }
+  if (amberContext) {
+    lines.push(
+      `## AMBER Capability Summary`,
+      `Total capabilities: ${amberContext.capabilities.length}`,
+      `Drifted: ${amberContext.driftedCapabilities}`,
+      `Tagged files: ${amberContext.taggedFiles} of ${amberContext.totalFiles} (${amberContext.taggedPercent}%)`,
+      ``,
+      amberContext.summary,
+      ``
+    );
+  }
+  return lines.join("\n");
+}
+function buildSystemPrompt12(audience) {
+  const audienceInstructions = audience === "executive" ? "You are writing for a CTO, VP Engineering, or executive stakeholder. Use ZERO technical jargon. Focus exclusively on business impact: hours saved, risk reduced, team health, delivery confidence. Translate engineering events into business outcomes." : "You are writing for a senior engineering team. Be specific \u2014 include file counts, capability names, cycle details, and concrete numbers. Engineers appreciate precision and honesty.";
+  return [
+    `You are Architecture Radio \u2014 the daily briefing that turns raw engineering signal into clear, compelling communication.`,
+    audienceInstructions,
+    `Keep Slack version punchy and action-oriented (~150 words). Email version has proper structure with greeting, paragraphs, and CTA (~400 words). Twitter version is a shareable insight \u2014 public-safe, no sensitive data, \u2264280 chars.`,
+    `Generate ONLY valid JSON \u2014 no markdown fences, no prose, no commentary outside the JSON object.`
+  ].join(" ");
+}
+function buildUserPrompt12(opts) {
+  const context = buildContext4(opts);
+  return [
+    `Generate today's Architecture Radio briefing for: ${opts.projectName}`,
+    `Audience: ${opts.audience}`,
+    ``,
+    context,
+    ``,
+    `## Output Format`,
+    `Return a single JSON object with this exact shape:`,
+    `{`,
+    `  "headline": "<1 powerful sentence capturing the most important thing \u2014 not a title, a statement>",`,
+    `  "slackVersion": "<~150 word Slack mrkdwn message \u2014 punchy, action-oriented, uses *bold*, _italic_, \u2022 bullets>",`,
+    `  "emailVersion": "<~400 word complete HTML email \u2014 full <html><head><body> with inline CSS, dark background, proper greeting, body paragraphs, clear CTA>",`,
+    `  "twitterVersion": "<insight worth sharing publicly \u2014 \u2264280 chars, no sensitive data, no project-specific names unless generic>"`,
+    `}`,
+    ``,
+    `Rules:`,
+    `- headline: not a generic recap \u2014 capture what MATTERS most today`,
+    `- slackVersion: start with the headline, use mrkdwn formatting`,
+    `- emailVersion: full HTML with inline styles, dark-mode friendly (#0f172a background), greeting, 2-3 paragraphs, bold CTA at end`,
+    `- twitterVersion: shareable engineering insight, safe for public, abstract enough for any team`,
+    `- ${opts.audience === "executive" ? "No technical jargon anywhere \u2014 business language only" : "Be technically precise \u2014 engineers hate vague"}`,
+    `- Derive all content from the digest context \u2014 no invented facts`,
+    `- Return ONLY valid JSON. No markdown fences. No prose outside the JSON.`
+  ].join("\n");
+}
+function buildFallback6(opts) {
+  const { digestContext: d, projectName, audience } = opts;
+  const isExec = audience === "executive";
+  const headline = d.newRisks.length > 0 ? isExec ? `${projectName} has ${d.newRisks.length} new architectural risk${d.newRisks.length > 1 ? "s" : ""} requiring attention.` : `${d.newDrifts.length} capability drift${d.newDrifts.length !== 1 ? "s" : ""} detected across ${d.filesChanged} changed files \u2014 action needed.` : isExec ? `${projectName} delivered a healthy week with ${d.commitCount} changes and stable architecture.` : `${d.commitCount} commits, ${d.filesChanged} files \u2014 architecture grade: ${d.grade ?? "N/A"}. ${d.healthSummary}`;
+  const slackVersion = [
+    `*\u{1F4E1} Architecture Radio \u2014 ${d.period}*`,
+    ``,
+    `*${headline}*`,
+    ``,
+    `\u2022 *Health:* ${d.healthSummary}`,
+    d.commitCount > 0 ? `\u2022 *Commits:* ${d.commitCount} across ${d.filesChanged} files` : "",
+    d.newDrifts.length > 0 ? `\u2022 *Drifts:* ${d.newDrifts.slice(0, 3).join(", ")}` : "",
+    d.newRisks.length > 0 ? `\u2022 *Risks:* ${d.newRisks.slice(0, 2).join(", ")}` : "",
+    ``,
+    `_Generated by forge0x2B_`
+  ].filter(Boolean).join("\n");
+  const emailVersion = `<!DOCTYPE html>
+<html lang="en">
+<head><meta charset="UTF-8"><meta name="viewport" content="width=device-width,initial-scale=1.0"><title>Architecture Radio \u2014 ${projectName}</title></head>
+<body style="margin:0;padding:0;background:#0f172a;font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',sans-serif">
+  <table width="100%" cellpadding="0" cellspacing="0" border="0" style="max-width:600px;margin:0 auto">
+    <tr><td style="padding:32px 32px 16px;border-bottom:1px solid #1e293b">
+      <div style="font-size:11px;color:#f97316;font-weight:600;letter-spacing:0.1em;text-transform:uppercase;margin-bottom:8px">\u{1F4E1} Architecture Radio</div>
+      <h1 style="margin:0;font-size:22px;font-weight:700;color:#f1f5f9;line-height:1.3">${headline}</h1>
+      <p style="margin:8px 0 0;color:#64748b;font-size:13px">${d.period}</p>
+    </td></tr>
+    <tr><td style="padding:24px 32px">
+      <p style="margin:0 0 16px;color:#94a3b8;font-size:14px;line-height:1.6">Hello,</p>
+      <p style="margin:0 0 16px;color:#cbd5e1;font-size:14px;line-height:1.6">${d.healthSummary} The team made ${d.commitCount} commit${d.commitCount !== 1 ? "s" : ""} across ${d.filesChanged} file${d.filesChanged !== 1 ? "s" : ""} during ${d.period}.</p>
+      ${d.newDrifts.length > 0 ? `<p style="margin:0 0 16px;color:#cbd5e1;font-size:14px;line-height:1.6"><strong style="color:#f1f5f9">Capability attention needed:</strong> ${d.newDrifts.slice(0, 3).join(", ")}.</p>` : ""}
+      ${d.newRisks.length > 0 ? `<p style="margin:0 0 16px;color:#cbd5e1;font-size:14px;line-height:1.6"><strong style="color:#fca5a5">New risks identified:</strong> ${d.newRisks.slice(0, 3).join(", ")}.</p>` : ""}
+      <p style="margin:24px 0 0;color:#64748b;font-size:12px;border-top:1px solid #1e293b;padding-top:16px">Generated with <a href="https://forge0x2b.dev" style="color:#f97316;text-decoration:none">forge0x2B</a> \xB7 Architecture intelligence for engineering teams</p>
+    </td></tr>
+  </table>
+</body>
+</html>`;
+  const twitterVersion = d.scoreDelta !== null ? `Engineering teams that track architectural drift resolve issues ${Math.abs(d.scoreDelta)}x faster. What gets measured gets fixed. \u{1F3D7}\uFE0F` : `The best architecture documentation is one your team actually reads \u2014 and it updates itself. \u{1F4E1}`;
+  return { headline, slackVersion, emailVersion, twitterVersion };
+}
+function parseLlmResponse6(raw) {
+  let text = raw.trim();
+  if (text.startsWith("```")) {
+    text = text.replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "").trim();
+  }
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  const getString = (key) => typeof parsed[key] === "string" ? parsed[key] : "";
+  const headline = getString("headline");
+  if (!headline) return null;
+  return {
+    headline,
+    slackVersion: getString("slackVersion"),
+    emailVersion: getString("emailVersion"),
+    twitterVersion: getString("twitterVersion").slice(0, 280)
+  };
+}
+async function generateRadio(options) {
+  try {
+    const response = await options.llm.complete({
+      systemPrompt: buildSystemPrompt12(options.audience),
+      messages: [{ role: "user", content: buildUserPrompt12(options) }],
+      maxTokens: 4096
+    });
+    const parsed = parseLlmResponse6(response.content);
+    if (parsed) return parsed;
+  } catch {
+  }
+  return buildFallback6(options);
+}
+
+// src/generators/generateArc42.ts
+var ARC42_ATTRIBUTION = `arc42 template \xA9 arc42.org, Creative Commons Attribution 4.0 International (CC-BY 4.0).
+See https://arc42.org for the original template and documentation.`;
+function buildSection5(opts) {
+  const { amberContext, projectName } = opts;
+  if (!amberContext || amberContext.capabilities.length === 0) {
+    return `## 5. Building Block View
+
+No AMBER capability data available. Run an AMBER scan to populate this section.
+
+*This section would show the top-level decomposition of the system into building blocks (capabilities) and their relationships.*
+`;
+  }
+  const lines = [
+    `## 5. Building Block View`,
+    ``,
+    `*Source: PRISM AMBER capability registry*`,
+    ``,
+    `### Level 1 \u2014 ${projectName} (Whitebox)`,
+    ``,
+    `**Total capabilities:** ${amberContext.capabilities.length}`,
+    `**Tagged files:** ${amberContext.taggedFiles} of ${amberContext.totalFiles} (${amberContext.taggedPercent.toFixed(0)}%)`,
+    ``,
+    `| Capability | Criticality | Lifecycle | Files |`,
+    `|------------|-------------|-----------|-------|`
+  ];
+  for (const cap of amberContext.capabilities) {
+    lines.push(
+      `| ${cap.name} | ${cap.criticality} | ${cap.lifecycle} | ${cap.files.length} |`
+    );
+  }
+  if (amberContext.driftedCapabilities > 0) {
+    lines.push(
+      ``,
+      `> **Note:** ${amberContext.driftedCapabilities} capabilities have documentation drift (files modified since last doc update).`
+    );
+  }
+  if (amberContext.orphanedFiles.length > 0) {
+    lines.push(
+      ``,
+      `### Orphaned Files (not assigned to any capability)`,
+      ``,
+      ...amberContext.orphanedFiles.slice(0, 10).map((f) => `- \`${f}\``),
+      amberContext.orphanedFiles.length > 10 ? `- *... and ${amberContext.orphanedFiles.length - 10} more*` : ``
+    );
+  }
+  return lines.join("\n");
+}
+function buildSection9(opts) {
+  const { digestContext: d } = opts;
+  const lines = [
+    `## 9. Architecture Decisions`,
+    ``,
+    `*Source: PRISM architecture signals*`,
+    ``
+  ];
+  if (d.newRisks.length > 0) {
+    lines.push(
+      `### Recent Architectural Risks Surfaced`,
+      ``,
+      ...d.newRisks.map((r) => `- ${r}`),
+      ``
+    );
+  }
+  if (d.newDrifts.length > 0 || d.resolvedDrifts.length > 0) {
+    lines.push(
+      `### Documentation State`,
+      ``
+    );
+    if (d.newDrifts.length > 0) {
+      lines.push(
+        `**New capability drifts** (${d.newDrifts.length}):`,
+        ...d.newDrifts.map((c) => `- ${c}`),
+        ``
+      );
+    }
+    if (d.resolvedDrifts.length > 0) {
+      lines.push(
+        `**Resolved drifts** (${d.resolvedDrifts.length}):`,
+        ...d.resolvedDrifts.map((c) => `- \u2713 ${c}`),
+        ``
+      );
+    }
+  }
+  lines.push(
+    `### Placeholder`,
+    ``,
+    `*Add ADRs (Architecture Decision Records) here. Use the forge0x2B ADR generator to produce MADR-format decisions from your codebase.*`,
+    ``,
+    `| ID | Title | Status | Date |`,
+    `|----|-------|--------|------|`,
+    `| ADR-001 | *Example decision* | Proposed | ${(/* @__PURE__ */ new Date()).toISOString().slice(0, 10)} |`
+  );
+  return lines.join("\n");
+}
+function buildSection10(opts) {
+  const { digestContext: d } = opts;
+  const lines = [
+    `## 10. Quality Requirements`,
+    ``,
+    `*Source: PRISM coherence metrics*`,
+    ``,
+    `### Quality Tree`,
+    ``,
+    `| Quality Goal | Scenario | Priority | Current State |`,
+    `|--------------|----------|----------|---------------|`
+  ];
+  const grade = d.grade ?? "N/A";
+  const gradeColor = grade.startsWith("A") ? "\u2713" : grade.startsWith("B") ? "~" : "\u26A0";
+  lines.push(
+    `| Architecture Coherence | Architecture grade \u2265 B | High | ${gradeColor} ${grade} |`,
+    `| Documentation Coverage | All capabilities documented | Medium | ${d.newDrifts.length === 0 ? "\u2713 No drift" : `\u26A0 ${d.newDrifts.length} drifted`} |`,
+    `| Change Stability | Low churn per capability | Medium | ${d.filesChanged} files changed this period |`
+  );
+  if (d.scoreStart !== null && d.scoreEnd !== null) {
+    const delta = d.scoreDelta !== null ? d.scoreDelta > 0 ? `+${d.scoreDelta}` : String(d.scoreDelta) : "N/A";
+    lines.push(
+      ``,
+      `### Coherence Score Trend`,
+      ``,
+      `| Period | Start | End | Delta |`,
+      `|--------|-------|-----|-------|`,
+      `| ${d.period} | ${d.scoreStart} | ${d.scoreEnd} | ${delta} |`
+    );
+  }
+  return lines.join("\n");
+}
+function buildSection11(opts) {
+  const { digestContext: d, amberContext } = opts;
+  const lines = [
+    `## 11. Risks and Technical Debt`,
+    ``,
+    `*Source: PRISM risk signals*`,
+    ``,
+    `### Risk Register`,
+    ``,
+    `| Risk | Probability | Impact | Mitigation |`,
+    `|------|------------|--------|------------|`
+  ];
+  if (d.newRisks.length > 0) {
+    for (const risk of d.newRisks) {
+      lines.push(`| ${risk} | Medium | Medium | Review and address in next sprint |`);
+    }
+  } else {
+    lines.push(`| *No new risks detected in ${d.period}* | \u2014 | \u2014 | \u2014 |`);
+  }
+  if (amberContext && amberContext.driftedCapabilities > 0) {
+    lines.push(`| Documentation drift: ${amberContext.driftedCapabilities} capabilities | High | Low | Run \`prism sync\` |`);
+  }
+  if (d.newDrifts.length > 0) {
+    lines.push(
+      ``,
+      `### Technical Debt \u2014 Documentation Drift`,
+      ``,
+      `The following capabilities have drifted documentation (code changed since last doc update):`,
+      ``,
+      ...d.newDrifts.map((c) => `- \`${c}\``)
+    );
+  }
+  lines.push(
+    ``,
+    `### Health Summary`,
+    ``,
+    `> ${d.healthSummary}`
+  );
+  return lines.join("\n");
+}
+function buildSection12(opts) {
+  const { amberContext, projectName } = opts;
+  const lines = [
+    `## 12. Glossary`,
+    ``,
+    `*Source: PRISM AMBER capability registry (Ubiquitous Language)*`,
+    ``,
+    `| Term | Definition |`,
+    `|------|------------|`
+  ];
+  if (amberContext && amberContext.capabilities.length > 0) {
+    for (const cap of amberContext.capabilities) {
+      const desc = cap.description ? cap.description : `${cap.criticality} ${cap.lifecycle} capability in ${projectName}.`;
+      lines.push(`| **${cap.name}** | ${desc} |`);
+    }
+  } else {
+    lines.push(
+      `| *No capabilities registered yet* | Run an AMBER scan to populate the ubiquitous language. |`
+    );
+  }
+  lines.push(
+    ``,
+    `*Add domain-specific terms and acronyms below as the glossary evolves.*`
+  );
+  return lines.join("\n");
+}
+function buildSection2() {
+  return `## 2. Constraints
+
+*Fill in the constraints that apply to your architecture. These are non-negotiable boundaries.*
+
+### Technical Constraints
+
+| Constraint | Background |
+|------------|------------|
+| *[e.g. Must run on Kubernetes]* | *[reason]* |
+| *[e.g. Postgres only \u2014 no other DBs]* | *[reason]* |
+
+### Organizational Constraints
+
+| Constraint | Background |
+|------------|------------|
+| *[e.g. Team of N developers]* | *[reason]* |
+| *[e.g. 2-week sprints]* | *[reason]* |
+
+### Conventions
+
+| Convention | Background |
+|------------|------------|
+| *[e.g. TypeScript strict mode]* | *[reason]* |
+`;
+}
+function buildSection6() {
+  return `## 6. Runtime View
+
+*Describe the important runtime scenarios \u2014 how the system behaves at runtime for key use cases.*
+
+### Scenario 1: [Name the scenario]
+
+\`\`\`
+[Actor] \u2192 [Component A] \u2192 [Component B] \u2192 [External System]
+         \u2190\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500 [Response] \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500
+\`\`\`
+
+**Description:** *[Describe what happens step by step.]*
+
+### Scenario 2: [Name the scenario]
+
+*[Add more scenarios as needed.]*
+`;
+}
+function buildSection7() {
+  return `## 7. Deployment View
+
+*Describe the technical infrastructure \u2014 environments, nodes, and how the system is distributed.*
+
+### Infrastructure Level 1
+
+\`\`\`
+[Environment: Production]
+  \u2514\u2500\u2500 [Server / Cloud Region]
+        \u251C\u2500\u2500 [Application Tier]
+        \u251C\u2500\u2500 [Database Tier]
+        \u2514\u2500\u2500 [External Services]
+\`\`\`
+
+### Environments
+
+| Environment | Purpose | URL / Access |
+|-------------|---------|-------------|
+| Production | Live system | *[URL]* |
+| Staging | Pre-production testing | *[URL]* |
+| Development | Local dev | localhost |
+
+*Fill in your actual infrastructure topology.*
+`;
+}
+function buildLlmContext(opts) {
+  const { projectName, projectDescription, teamSize, techStack, digestContext: d, amberContext } = opts;
+  const lines = [
+    `Project: ${projectName}`,
+    projectDescription ? `Description: ${projectDescription}` : "",
+    teamSize ? `Team size: ${teamSize}` : "",
+    techStack && techStack.length > 0 ? `Tech stack: ${techStack.join(", ")}` : "",
+    ``,
+    `Architecture health: ${d.healthSummary}`,
+    d.grade ? `Grade: ${d.grade}` : "",
+    `Period: ${d.period}`,
+    `Commits: ${d.commitCount}`,
+    `Files changed: ${d.filesChanged}`,
+    d.topCommits.length > 0 ? `Recent commits: ${d.topCommits.slice(0, 4).join("; ")}` : ""
+  ].filter(Boolean);
+  if (amberContext && amberContext.capabilities.length > 0) {
+    lines.push(
+      ``,
+      `Capabilities: ${amberContext.capabilities.map((c) => c.name).join(", ")}`,
+      `Total capabilities: ${amberContext.capabilities.length}`
+    );
+  }
+  return lines.join("\n");
+}
+async function generateLlmSections(opts) {
+  const context = buildLlmContext(opts);
+  const systemPrompt = `You are an expert software architect generating arc42 architecture documentation.
+Write concise, informative Markdown content for each section.
+Base your content on the provided PRISM architecture data and project information.
+Use professional, clear language appropriate for technical documentation.
+Do not reproduce copyrighted content. All insights are derived from the provided data.`;
+  const userPrompt = `Generate four arc42 documentation sections for this project.
+
+Project data:
+${context}
+
+Generate the following four sections in Markdown. Each section should start with the exact header shown.
+Keep each section focused and practical \u2014 100\u2013300 words per section.
+
+---
+## 1. Introduction and Goals
+
+Write an overview of:
+- What the system does (based on the project name, description, and capabilities)
+- Key quality goals (2\u20133 measurable goals)
+- Key stakeholders and their expectations
+
+---
+## 3. Context and Scope
+
+Write:
+- A system context description (what the system does, what it connects to)
+- External systems and actors that interact with the system
+- What is explicitly out of scope
+
+---
+## 4. Solution Strategy
+
+Write:
+- Core technology decisions and why
+- Key architectural patterns used (infer from the tech stack and capability structure)
+- How quality goals are addressed by the architecture
+
+---
+## 8. Crosscutting Concepts
+
+Identify crosscutting concerns from the capability names and architecture:
+- Security / authentication approach
+- Error handling and logging
+- Caching strategy (if applicable)
+- Testing approach
+- Any other crosscutting patterns visible from the capability structure
+`;
+  const response = await opts.llm.complete({
+    systemPrompt,
+    messages: [{ role: "user", content: userPrompt }],
+    maxTokens: 4096
+  });
+  const text = response.content;
+  function extractSection(marker, nextMarker) {
+    const start = text.indexOf(marker);
+    if (start === -1) return "";
+    const end = nextMarker ? text.indexOf(nextMarker, start + marker.length) : text.length;
+    return end === -1 ? text.slice(start).trim() : text.slice(start, end).trim();
+  }
+  const s1 = extractSection("## 1. Introduction", "## 3. Context");
+  const s3 = extractSection("## 3. Context", "## 4. Solution");
+  const s4 = extractSection("## 4. Solution", "## 8. Crosscutting");
+  const s8 = extractSection("## 8. Crosscutting", "");
+  return { s1, s3, s4, s8 };
+}
+function fallbackSection1(opts) {
+  const { projectName, projectDescription, amberContext } = opts;
+  const capCount = amberContext?.capabilities.length ?? 0;
+  return `## 1. Introduction and Goals
+
+**${projectName}** ${projectDescription ?? "is a software system documented using arc42."}
+
+### Quality Goals
+
+| Priority | Quality Goal | Scenario |
+|----------|-------------|----------|
+| 1 | Correctness | System produces correct results |
+| 2 | Maintainability | New features can be added without regression |
+| 3 | Performance | System responds within acceptable time |
+
+### Stakeholders
+
+| Role | Expectations |
+|------|-------------|
+| Development Team | Clear architecture structure, ${capCount} documented capabilities |
+| Product Owner | Feature delivery aligned with roadmap |
+| Operations | System is deployable and observable |
+
+*Generated from PRISM data \u2014 enrich with project-specific goals.*
+`;
+}
+function fallbackSection3(opts) {
+  const { projectName } = opts;
+  return `## 3. Context and Scope
+
+### System Context
+
+**${projectName}** interacts with the following external systems:
+
+| External System | Description | Relationship |
+|----------------|-------------|-------------|
+| *[Users / Clients]* | End users of the system | Consumer |
+| *[External API / Service]* | *[Describe]* | Integration |
+| *[Database]* | Data persistence | Provider |
+
+### In Scope
+
+*[Describe what the system is responsible for.]*
+
+### Out of Scope
+
+*[Describe what is deliberately excluded.]*
+`;
+}
+function fallbackSection4(opts) {
+  const { techStack } = opts;
+  const stack = techStack && techStack.length > 0 ? techStack.join(", ") : "TypeScript, Next.js";
+  return `## 4. Solution Strategy
+
+### Technology Decisions
+
+**Tech stack:** ${stack}
+
+| Decision | Rationale |
+|----------|-----------|
+| *[Framework choice]* | *[Why this framework]* |
+| *[Database choice]* | *[Why this DB]* |
+| *[Architecture pattern]* | *[Why this pattern]* |
+
+### Architectural Approach
+
+*Describe the core architectural approach \u2014 e.g. monolith, microservices, event-driven, CQRS.*
+
+### Quality Goal Mapping
+
+| Quality Goal | Architecture Response |
+|-------------|----------------------|
+| Correctness | *[How architecture ensures correctness]* |
+| Maintainability | *[How architecture supports maintainability]* |
+`;
+}
+function fallbackSection8(opts) {
+  return `## 8. Crosscutting Concepts
+
+### Security
+
+*Describe the authentication and authorization approach.*
+
+### Error Handling
+
+*Describe the error handling and logging strategy.*
+
+### Observability
+
+*Describe how the system is monitored \u2014 logs, metrics, tracing.*
+
+### Testing Strategy
+
+*Describe the testing levels \u2014 unit, integration, end-to-end.*
+
+### Configuration
+
+*Describe how the system is configured across environments.*
+`;
+}
+async function generateArc42(opts) {
+  const generatedAt = (/* @__PURE__ */ new Date()).toISOString();
+  const s5Content = buildSection5(opts);
+  const s9Content = buildSection9(opts);
+  const s10Content = buildSection10(opts);
+  const s11Content = buildSection11(opts);
+  const s12Content = buildSection12(opts);
+  const s2Content = buildSection2();
+  const s6Content = buildSection6();
+  const s7Content = buildSection7();
+  let s1Content;
+  let s3Content;
+  let s4Content;
+  let s8Content;
+  try {
+    const llmSections = await generateLlmSections(opts);
+    s1Content = llmSections.s1 || fallbackSection1(opts);
+    s3Content = llmSections.s3 || fallbackSection3(opts);
+    s4Content = llmSections.s4 || fallbackSection4(opts);
+    s8Content = llmSections.s8 || fallbackSection8(opts);
+  } catch {
+    s1Content = fallbackSection1(opts);
+    s3Content = fallbackSection3(opts);
+    s4Content = fallbackSection4(opts);
+    s8Content = fallbackSection8();
+  }
+  const sections = [
+    { number: 1, title: "Introduction and Goals", content: s1Content, dataSource: "llm" },
+    { number: 2, title: "Constraints", content: s2Content, dataSource: "template" },
+    { number: 3, title: "Context and Scope", content: s3Content, dataSource: "llm" },
+    { number: 4, title: "Solution Strategy", content: s4Content, dataSource: "llm" },
+    { number: 5, title: "Building Block View", content: s5Content, dataSource: "prism" },
+    { number: 6, title: "Runtime View", content: s6Content, dataSource: "template" },
+    { number: 7, title: "Deployment View", content: s7Content, dataSource: "template" },
+    { number: 8, title: "Crosscutting Concepts", content: s8Content, dataSource: "llm" },
+    { number: 9, title: "Architecture Decisions", content: s9Content, dataSource: "prism" },
+    { number: 10, title: "Quality Requirements", content: s10Content, dataSource: "prism" },
+    { number: 11, title: "Risks and Technical Debt", content: s11Content, dataSource: "prism" },
+    { number: 12, title: "Glossary", content: s12Content, dataSource: "prism" }
+  ];
+  const fullMarkdown = [
+    `<!-- This document uses the arc42 template \xA9 arc42.org, licensed under CC-BY 4.0 -->`,
+    `<!-- Generated by PRISM0x2A + forge0x2B \u2014 Architecture intelligence -->`,
+    `<!-- ${generatedAt} -->`,
+    ``,
+    `# Architecture Documentation \u2014 ${opts.projectName}`,
+    ``,
+    `> ${ARC42_ATTRIBUTION}`,
+    ``,
+    ...sections.map((s) => s.content)
+  ].join("\n\n");
+  return {
+    sections,
+    fullMarkdown,
+    attribution: ARC42_ATTRIBUTION,
+    generatedAt
+  };
+}
+
+// src/generators/generateKnowledgeCapture.ts
+function buildContext5(opts) {
+  const { digestContext: d, amberContext, projectName, departingDeveloper, focusCapabilities } = opts;
+  const lines = [
+    `## Knowledge Capture Context`,
+    `Project: ${projectName}`,
+    departingDeveloper ? `Departing Developer: ${departingDeveloper}` : `Departing Developer: (not specified \u2014 write for general knowledge transfer)`,
+    focusCapabilities && focusCapabilities.length > 0 ? `Focus Capabilities: ${focusCapabilities.join(", ")}` : `Focus Capabilities: all`,
+    ``,
+    `## Recent Activity \u2014 ${d.period}`,
+    `Commits: ${d.commitCount}`,
+    `Files changed: ${d.filesChanged}`,
+    `Architecture grade: ${d.grade ?? "N/A"}`,
+    `Health: ${d.healthSummary}`,
+    ``
+  ];
+  if (d.topCommits.length > 0) {
+    lines.push(`### Recent Commits`, ...d.topCommits.map((c) => `- ${c}`), ``);
+  }
+  if (d.newDrifts.length > 0) {
+    lines.push(
+      `### Drifted Capabilities (need documentation attention)`,
+      ...d.newDrifts.map((c) => `- ${c}`),
+      ``
+    );
+  }
+  if (d.newRisks.length > 0) {
+    lines.push(`### Known Risks`, ...d.newRisks.map((r) => `- ${r}`), ``);
+  }
+  if (amberContext) {
+    lines.push(`## AMBER Capability Registry`);
+    lines.push(`Total capabilities: ${amberContext.capabilities.length}`);
+    lines.push(`Drifted: ${amberContext.driftedCapabilities}`);
+    lines.push(`Tagged files: ${amberContext.taggedFiles} / ${amberContext.totalFiles} (${amberContext.taggedPercent}%)`);
+    lines.push(``);
+    const toInclude = focusCapabilities && focusCapabilities.length > 0 ? amberContext.capabilities.filter(
+      (c) => focusCapabilities.some((f) => c.name.toLowerCase().includes(f.toLowerCase()) || c.id.toLowerCase().includes(f.toLowerCase()))
+    ) : amberContext.capabilities;
+    if (toInclude.length > 0) {
+      lines.push(`### Capabilities`);
+      for (const cap of toInclude.slice(0, 20)) {
+        lines.push(`#### ${cap.name} (${cap.id})`);
+        lines.push(`- Criticality: ${cap.criticality}`);
+        lines.push(`- Lifecycle: ${cap.lifecycle}`);
+        lines.push(`- Files: ${cap.files.length}`);
+        lines.push(`- Drift: ${cap.driftCount} file${cap.driftCount !== 1 ? "s" : ""} out of date`);
+        if (cap.description) lines.push(`- Description: ${cap.description}`);
+        lines.push(``);
       }
-      return;
     }
-    if (visited.has(node)) return;
-    visited.add(node);
-    stack.add(node);
-    for (const neighbor of graph.get(node) ?? []) {
-      dfs(neighbor, [...path, node]);
+    if (amberContext.orphanedFiles.length > 0) {
+      lines.push(
+        `### Orphaned Files (no capability tag)`,
+        ...amberContext.orphanedFiles.slice(0, 10).map((f) => `- ${f}`),
+        ``
+      );
     }
-    stack.delete(node);
-  }
-  for (const node of graph.keys()) {
-    if (!visited.has(node)) dfs(node, []);
   }
-  return cycles;
+  return lines.join("\n");
 }
-function buildUserPrompt5(blueprint, opts) {
-  const tone = opts.tone ?? "analytical";
-  const length = opts.length ?? "medium";
-  const format = opts.format ?? "markdown";
-  const hotFiles = blueprint.files.slice().sort((a, b) => (b.importedByCount ?? 0) + (b.importCount ?? 0) - ((a.importedByCount ?? 0) + (a.importCount ?? 0))).slice(0, 10);
-  const highFanOut = blueprint.files.slice().sort((a, b) => (b.importCount ?? 0) - (a.importCount ?? 0)).slice(0, 8);
-  const importCycles = detectImportCycles(blueprint.edges);
-  const lines = [
-    `Generate a refactoring report in ${format} format for the codebase at "${blueprint.targetPath}".`,
-    `Tone: ${tone}. Length: ${length}.`,
+function buildSystemPrompt13(opts) {
+  const devName = opts.departingDeveloper ?? "a departing developer";
+  return [
+    `You are generating a Knowledge Capture document \u2014 a complete architectural brain dump before ${devName} leaves the team.`,
+    `This document is crucial. It preserves institutional knowledge that would otherwise walk out the door.`,
+    `Write with warmth and humanity \u2014 this is someone's legacy. Be thorough, specific, and genuinely useful.`,
+    `Make it feel like ${devName} wrote it themselves, with care for the next person who will maintain this system.`,
+    `Generate ONLY valid JSON \u2014 no markdown fences, no prose, no commentary outside the JSON object.`
+  ].join(" ");
+}
+function buildUserPrompt13(opts) {
+  const context = buildContext5(opts);
+  const devName = opts.departingDeveloper ?? "the departing developer";
+  return [
+    `Generate a complete Knowledge Capture document for: ${opts.projectName}`,
+    `This preserves ${devName}'s architectural knowledge for the team.`,
     ``,
-    `## Architecture Metrics`,
-    `- Total files: ${blueprint.stats.totalFiles}`,
-    `- Dependency edges: ${blueprint.stats.runtimeEdges}`,
-    `- Categories: ${JSON.stringify(blueprint.categories)}`,
+    context,
     ``,
-    `## High-Coupling Hot Spots (high total connections)`,
-    ...hotFiles.map((f) => `- ${f.path} \u2014 in: ${f.importedByCount ?? 0}, out: ${f.importCount ?? 0}, lines: ${f.lineCount ?? "?"}`),
+    `## Required Document Structure`,
+    `The document must cover these 6 sections:`,
+    `1. Architecture Overview \u2014 what the system does, how capabilities relate to each other, the mental model`,
+    `2. Critical Capabilities \u2014 deep dive on the most important/drifted/complex capabilities, what makes them tricky`,
+    `3. Known Issues & Workarounds \u2014 things that don't work as expected, temporary hacks, "do not touch" zones`,
+    `4. Where to Start \u2014 recommended reading order for a new developer, which files to read first`,
+    `5. Tribal Knowledge \u2014 things not obvious from code: why decisions were made, what was tried and failed, hidden dependencies`,
+    `6. Onboarding Checklist \u2014 concrete 10-step list for a new team member joining this project`,
     ``,
-    `## High Fan-Out Files (many outgoing dependencies)`,
-    ...highFanOut.map((f) => `- ${f.path} \u2014 imports: ${f.importCount ?? 0}`),
+    `## Output Format`,
+    `Return a single JSON object with this exact shape:`,
+    `{`,
+    `  "sections": [`,
+    `    { "title": "<section title>", "content": "<full markdown content for this section, multiple paragraphs>" }`,
+    `  ],`,
+    `  "criticalKnowledge": ["<most important thing 1>", "<most important thing 2>", ...],`,
+    `  "onboardingChecklist": ["<step 1: specific action>", "<step 2: specific action>", ...]`,
+    `}`,
     ``,
-    `## Import Cycles Detected (${importCycles.length})`,
-    ...importCycles.length > 0 ? importCycles.map((c) => `- ${c}`) : ["No cycles detected in sampled edges"],
+    `Rules:`,
+    `- sections: exactly 6, in the order above, with rich content (not just bullets)`,
+    `- criticalKnowledge: 5-8 bullets \u2014 the things a new developer MUST know or they will make mistakes`,
+    `- onboardingChecklist: exactly 10 items \u2014 concrete, actionable steps (not "read the README", but "read src/auth/ starting with session.ts")`,
+    `- Write as if ${devName} is speaking directly to their replacement \u2014 warm, honest, specific`,
+    `- Include file paths and capability names from the context where relevant`,
+    `- Derive all content from the context \u2014 no invented facts`,
+    `- Return ONLY valid JSON. No markdown fences. No prose outside the JSON.`
+  ].join("\n");
+}
+function sectionsToMarkdown(sections, opts, criticalKnowledge, onboardingChecklist) {
+  const devName = opts.departingDeveloper ?? "the team";
+  const date = (/* @__PURE__ */ new Date()).toLocaleDateString("en-US", { year: "numeric", month: "long", day: "numeric" });
+  const parts = [
+    `# Knowledge Capture: ${opts.projectName}`,
     ``,
-    `Identify refactoring priorities: coupling issues, over-large files, circular dependencies, layer violations. Suggest concrete refactoring actions with rationale.`
+    `> **Prepared by:** ${devName}  `,
+    `> **Date:** ${date}  `,
+    `> **Purpose:** Preserve architectural knowledge for the next maintainer`,
+    ``,
+    `---`,
+    ``,
+    `## TL;DR \u2014 Critical Knowledge`,
+    ``,
+    ...criticalKnowledge.map((k) => `- ${k}`),
+    ``,
+    `---`,
+    ``
   ];
-  return lines.join("\n");
-}
-async function generateRefactoringReport(blueprint, opts, provider) {
-  if (!blueprint) {
-    return {
-      text: NO_DATA_MSG4,
-      metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: 0, generator: "forgesmith" }
-    };
+  for (const section of sections) {
+    parts.push(`## ${section.title}`, ``, section.content, ``, `---`, ``);
   }
-  const response = await provider.complete({
-    systemPrompt: buildSystemPrompt5(opts.tone ?? "analytical"),
-    messages: [{ role: "user", content: buildUserPrompt5(blueprint, opts) }],
-    maxTokens: 3072
-  });
-  return {
-    text: response.content,
-    metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: response.usedTokens, generator: "forgesmith" }
-  };
-}
-
-// src/generators/askDrivenAsset.ts
-var NO_DATA_MSG5 = "No Blueprint data available. Run prism scan first.";
-var FORMAT_GUIDES = {
-  markdown: { name: "Markdown document", structure: "Use headers, bullet lists, and code blocks where appropriate.", maxTokens: 3072 },
-  blog: { name: "blog post", structure: "Write with an engaging intro, clear sections, a conclusion, and a call-to-action.", maxTokens: 3072 },
-  social: { name: "social media post", structure: "Write concise, punchy content suitable for Twitter/LinkedIn. Max 280 characters for Twitter mode.", maxTokens: 512 },
-  email: { name: "email", structure: "Use Subject:, greeting, body paragraphs, and a sign-off.", maxTokens: 1024 },
-  slack: { name: "Slack message", structure: "Keep it conversational, use *bold* for emphasis, bullet points for lists. Max 3 paragraphs.", maxTokens: 512 },
-  slide: { name: "presentation outline", structure: "Structure as slide titles with 3-5 bullet points each. Include a title slide and summary slide.", maxTokens: 2048 }
-};
-function buildSystemPrompt6(format, tone) {
-  const toneMap = {
-    professional: "You are a professional technical writer and developer advocate.",
-    casual: "You are a friendly engineering blogger who writes in an approachable, conversational style.",
-    technical: "You are a senior software engineer writing precise, implementation-focused content.",
-    executive: "You are a VP of Engineering writing high-level, business-value-focused content for leadership."
-  };
-  const guide = FORMAT_GUIDES[format];
-  return `${toneMap[tone] ?? toneMap.professional} Generate a ${guide.name} based on the user's question and the provided codebase architecture context. ${guide.structure} Write only the requested content \u2014 no preamble, no meta-commentary.`;
-}
-function buildUserPrompt6(blueprint, question, opts) {
-  const format = opts.format ?? "markdown";
-  const tone = opts.tone ?? "professional";
-  const length = opts.length ?? "medium";
-  const guide = FORMAT_GUIDES[format];
-  const lengthGuide = {
-    short: "Keep it concise \u2014 1-2 paragraphs or equivalent.",
-    medium: "Medium length \u2014 3-5 paragraphs or equivalent.",
-    long: "Detailed and comprehensive \u2014 cover the topic thoroughly."
-  }[length];
-  const topFiles = blueprint.files.slice().sort((a, b) => (b.importedByCount ?? 0) - (a.importedByCount ?? 0)).slice(0, 12);
-  const lines = [
-    `## User's Question`,
-    question,
+  parts.push(
+    `## Onboarding Checklist`,
     ``,
-    `## Output Requirements`,
-    `- Format: ${guide.name}`,
-    `- Tone: ${tone}`,
-    `- Length: ${length} (${lengthGuide})`,
+    `For the next person joining this project:`,
     ``,
-    `## Codebase Architecture Context`,
-    `Target: ${blueprint.targetPath}`,
-    `Total files: ${blueprint.stats.totalFiles} | Dependency edges: ${blueprint.stats.runtimeEdges}`,
-    `Categories: app=${blueprint.categories.app ?? 0}, components=${blueprint.categories.component ?? 0}, lib=${blueprint.categories.lib ?? 0}, hooks=${blueprint.categories.hook ?? 0}`,
+    ...onboardingChecklist.map((step, i) => `- [ ] **Step ${i + 1}:** ${step}`),
     ``,
-    `Key files (by usage):`,
-    ...topFiles.map((f) => `- ${f.path} [${f.category ?? "?"}] \u2014 imported by ${f.importedByCount ?? 0} files, ${f.lineCount ?? "?"} lines`)
+    `---`,
+    ``,
+    `*Generated with [forge0x2B](https://forge0x2b.dev) \xB7 Architecture intelligence for engineering teams*`
+  );
+  return parts.join("\n");
+}
+function buildFallback7(opts) {
+  const { digestContext: d, amberContext, projectName, departingDeveloper } = opts;
+  const critCaps = amberContext?.capabilities.filter((c) => c.criticality === "critical" || c.driftCount > 0).slice(0, 5).map((c) => c.name) ?? [];
+  const sections = [
+    {
+      title: "Architecture Overview",
+      content: `${projectName} is composed of ${amberContext?.capabilities.length ?? "several"} tracked capabilities. ${d.healthSummary}
+
+The system currently has ${d.filesChanged} recently changed files and a coherence grade of ${d.grade ?? "unknown"}.`
+    },
+    {
+      title: "Critical Capabilities",
+      content: critCaps.length > 0 ? `The following capabilities require the most attention:
+
+${critCaps.map((c) => `**${c}** \u2014 review all files tagged with this capability`).join("\n\n")}` : `Review the AMBER capability registry for the current capability map and any drift indicators.`
+    },
+    {
+      title: "Known Issues & Workarounds",
+      content: d.newRisks.length > 0 ? `The following risks have been identified:
+
+${d.newRisks.map((r) => `- **${r}**: Review related files before making changes`).join("\n")}` : `No critical known issues at time of writing. Check the PRISM risk registry for the latest state.`
+    },
+    {
+      title: "Where to Start",
+      content: `Start by understanding the capability structure. Read \`.amber/capabilities.md\` for the full capability registry, then \`.amber/state.json\` for file-to-capability mappings.
+
+Focus on capabilities with high drift counts first \u2014 those are where documentation has fallen behind the code.`
+    },
+    {
+      title: "Tribal Knowledge",
+      content: `Key things not obvious from reading the code:
+
+- The AMBER tags in source files are the source of truth for capability ownership
+- Drift count indicates documentation debt \u2014 files changed without updating @amber-doc tags
+- Architecture score changes over time \u2014 check \`.green/\` for trend data`
+    },
+    {
+      title: "Architecture Decisions",
+      content: `${departingDeveloper ? `${departingDeveloper} made` : "Key"} architectural decisions are not yet fully documented. Recommended: schedule a knowledge transfer session and ask specifically about: capability boundaries, cross-capability dependencies, and any "do not touch" areas.`
+    }
   ];
-  if (blueprint.edges.length > 0) {
-    const edgeSample = blueprint.edges.slice(0, 15);
-    lines.push(``, `Dependency edges (sample):`, ...edgeSample.map((e) => `- ${e.from} \u2192 ${e.to}`));
-  }
-  lines.push(``, `Answer the user's question using the architecture context above. Generate the ${guide.name} now.`);
-  return lines.join("\n");
+  const criticalKnowledge = [
+    `The AMBER capability registry (.amber/) is the source of truth for how the codebase is organized`,
+    `Drift count > 0 means documentation hasn't kept pace with code changes \u2014 fix before adding features`,
+    amberContext ? `${amberContext.driftedCapabilities} of ${amberContext.capabilities.length} capabilities currently have drift` : `Check drift count before starting work`,
+    d.newRisks.length > 0 ? `${d.newRisks.length} active risk${d.newRisks.length > 1 ? "s" : ""} identified: ${d.newRisks.slice(0, 2).join(", ")}` : `Monitor architecture score for regression`,
+    `Architecture grade: ${d.grade ?? "unknown"} \u2014 understand what's driving this before making broad changes`
+  ].filter(Boolean);
+  const onboardingChecklist = [
+    `Read this entire document before writing any code`,
+    `Install and run PRISM to get a live architecture view`,
+    `Review .amber/capabilities.md \u2014 understand every capability and its criticality`,
+    `Check .amber/state.json to see which files belong to which capabilities`,
+    `Run the test suite and confirm it passes before making changes`,
+    `Read the top ${Math.min(d.topCommits.length || 5, 5)} recent commits to understand current momentum`,
+    `Review capabilities with drift count > 0 \u2014 read both the code and the docs`,
+    `Set up forge0x2B to get daily Architecture Radio briefings`,
+    `Ask the team about any "do not touch" areas that aren't documented`,
+    `Make your first PR small and well-scoped to verify your understanding`
+  ];
+  const markdownDoc = sectionsToMarkdown(sections, opts, criticalKnowledge, onboardingChecklist);
+  return { markdownDoc, sections, criticalKnowledge, onboardingChecklist };
 }
-async function generateAskDrivenAsset(blueprint, question, opts, provider) {
-  if (!blueprint) {
-    return {
-      text: NO_DATA_MSG5,
-      metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: 0, generator: "forgesmith" }
-    };
+function parseLlmResponse7(raw, opts) {
+  let text = raw.trim();
+  if (text.startsWith("```")) {
+    text = text.replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "").trim();
   }
-  if (!question || !question.trim()) {
+  let parsed;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  const rawSections = Array.isArray(parsed.sections) ? parsed.sections : [];
+  const sections = rawSections.filter((s) => typeof s.title === "string").map((s) => {
+    const item = s;
     return {
-      text: "No question provided. Please ask something about your codebase.",
-      metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: 0, generator: "forgesmith" }
+      title: item.title,
+      content: typeof item.content === "string" ? item.content : ""
     };
-  }
-  const format = opts.format ?? "markdown";
-  const tone = opts.tone ?? "professional";
-  const maxTokens = FORMAT_GUIDES[format].maxTokens;
-  const response = await provider.complete({
-    systemPrompt: buildSystemPrompt6(format, tone),
-    messages: [{ role: "user", content: buildUserPrompt6(blueprint, question, opts) }],
-    maxTokens
   });
-  return {
-    text: response.content,
-    metadata: { generatedAt: (/* @__PURE__ */ new Date()).toISOString(), usedTokens: response.usedTokens, generator: "forgesmith" }
-  };
+  if (sections.length === 0) return null;
+  const rawCritical = Array.isArray(parsed.criticalKnowledge) ? parsed.criticalKnowledge : [];
+  const criticalKnowledge = rawCritical.filter((k) => typeof k === "string").map((k) => k);
+  const rawChecklist = Array.isArray(parsed.onboardingChecklist) ? parsed.onboardingChecklist : [];
+  const onboardingChecklist = rawChecklist.filter((k) => typeof k === "string").map((k) => k);
+  const markdownDoc = sectionsToMarkdown(sections, opts, criticalKnowledge, onboardingChecklist);
+  return { markdownDoc, sections, criticalKnowledge, onboardingChecklist };
+}
+async function generateKnowledgeCapture(options) {
+  try {
+    const response = await options.llm.complete({
+      systemPrompt: buildSystemPrompt13(options),
+      messages: [{ role: "user", content: buildUserPrompt13(options) }],
+      maxTokens: 8192
+    });
+    const parsed = parseLlmResponse7(response.content, options);
+    if (parsed) return parsed;
+  } catch {
+  }
+  return buildFallback7(options);
 }
 
 // src/forge/types.ts
@@ -1905,6 +4327,80 @@ var WIDGET_TEMPLATE_SOCIAL_PROOF = {
     { id: "show_divider", label: "Show divider", type: "select", options: ["yes", "no"], default: "no" }
   ]
 };
+var WIDGET_TEMPLATE_ANIMATED_STAT = {
+  id: "animated-stat",
+  name: "Animated Stat",
+  description: "Full-card animated number counter. Perfect for LinkedIn/Instagram. Export as video or GIF.",
+  free_tier: false,
+  animated: true,
+  animDurationMs: 2200,
+  videoDurationMs: 3800,
+  defaultWidth: 600,
+  defaultHeight: 400,
+  exportFormats: ["standalone", "html", "markdown"],
+  slots: [
+    { id: "value", label: "Value (number)", type: "number", required: true, placeholder: "138", default: "138" },
+    { id: "label", label: "Label", type: "text", required: true, placeholder: "Files analysed", default: "Files analysed" },
+    { id: "unit", label: "Unit (optional)", type: "text", placeholder: "k LOC", default: "" },
+    { id: "sublabel", label: "Sublabel / context", type: "text", placeholder: "across 3 modules", default: "" },
+    { id: "delta", label: "Delta text", type: "text", placeholder: "+22% this sprint", default: "" },
+    { id: "delta_direction", label: "Delta direction", type: "select", options: ["up", "down", "neutral"], default: "up" }
+  ]
+};
+var WIDGET_TEMPLATE_RELEASE_CARD = {
+  id: "release-card",
+  name: "Release Card",
+  description: "Animated release announcement \u2014 version, headline, and staggered change list. Great for LinkedIn posts.",
+  free_tier: false,
+  animated: true,
+  animDurationMs: 1200,
+  videoDurationMs: 3e3,
+  defaultWidth: 600,
+  defaultHeight: 400,
+  exportFormats: ["standalone", "html", "markdown"],
+  slots: [
+    { id: "version", label: "Version", type: "text", required: true, placeholder: "v2.4.0", default: "v2.4.0" },
+    { id: "headline", label: "Headline", type: "text", required: true, placeholder: "Faster. Smarter. Leaner.", default: "Faster. Smarter. Leaner." },
+    { id: "changes", label: "Changes (one per line)", type: "multiline", required: true, placeholder: "40% faster cold starts\nNew Capability Registry\nFixed circular import detection", default: "40% faster cold starts\nNew Capability Registry\nFixed circular import detection" },
+    { id: "tag", label: "Release tag", type: "select", options: ["feat", "fix", "perf", "break", ""], default: "feat" }
+  ]
+};
+var WIDGET_TEMPLATE_CHART_BARS = {
+  id: "chart-bars",
+  name: "Chart \u2014 Bars",
+  description: "Animated Chart.js bar chart from your codebase metrics. Perfect for visualising churn, coverage, or LOC.",
+  free_tier: false,
+  animated: true,
+  animDurationMs: 1600,
+  videoDurationMs: 3200,
+  defaultWidth: 640,
+  defaultHeight: 420,
+  exportFormats: ["standalone", "html", "markdown"],
+  slots: [
+    { id: "title", label: "Chart title", type: "text", placeholder: "Files by module", default: "Files by module" },
+    { id: "labels", label: "Labels (one per line)", type: "multiline", required: true, placeholder: "api\nlib\ncomponents\napp", default: "api\nlib\ncomponents\napp" },
+    { id: "values", label: "Values (one per line)", type: "multiline", required: true, placeholder: "24\n18\n41\n12", default: "24\n18\n41\n12" },
+    { id: "orientation", label: "Orientation", type: "select", options: ["vertical", "horizontal"], default: "vertical" },
+    { id: "accent_color", label: "Bar colour (CSS)", type: "color", placeholder: "#f97316", default: "" }
+  ]
+};
+var WIDGET_TEMPLATE_ARCHITECTURE_BADGE = {
+  id: "architecture-badge",
+  name: "Architecture Badge",
+  description: "Animated tech-stack showcase with Lucide icons. Items pop in with spring animation \u2014 great for profile posts.",
+  free_tier: false,
+  animated: true,
+  animDurationMs: 900,
+  videoDurationMs: 2500,
+  defaultWidth: 580,
+  defaultHeight: 360,
+  exportFormats: ["standalone", "html", "markdown"],
+  slots: [
+    { id: "title", label: "Title", type: "text", placeholder: "Built with", default: "Built with" },
+    { id: "items", label: "Stack items (one per line)", type: "multiline", required: true, placeholder: "TypeScript\nNext.js\nPostgres\nTailwind", default: "TypeScript\nNext.js\nPostgres\nTailwind" },
+    { id: "subtitle", label: "Subtitle / tagline", type: "text", placeholder: "100% local-first \xB7 zero telemetry", default: "" }
+  ]
+};
 var BUNDLED_WIDGET_TEMPLATES = [
   WIDGET_TEMPLATE_STAT_CARD,
   WIDGET_TEMPLATE_FEATURE_GRID,
@@ -1913,7 +4409,12 @@ var BUNDLED_WIDGET_TEMPLATES = [
   WIDGET_TEMPLATE_METRIC_BADGE,
   WIDGET_TEMPLATE_PRICING_TIER,
   WIDGET_TEMPLATE_CHANGELOG_ROW,
-  WIDGET_TEMPLATE_SOCIAL_PROOF
+  WIDGET_TEMPLATE_SOCIAL_PROOF,
+  // Animated — standalone + video export
+  WIDGET_TEMPLATE_ANIMATED_STAT,
+  WIDGET_TEMPLATE_RELEASE_CARD,
+  WIDGET_TEMPLATE_CHART_BARS,
+  WIDGET_TEMPLATE_ARCHITECTURE_BADGE
 ];
 var FREE_TIER_WIDGET_IDS = BUNDLED_WIDGET_TEMPLATES.filter((t) => t.free_tier).map((t) => t.id);
 function getWidgetTemplate(id) {
@@ -2072,6 +4573,184 @@ ${divider ? `<hr style="border:none;border-top:1px solid rgba(255,255,255,.08);m
 ${logos.map((l) => `  <span style="font-weight:600;font-size:.875rem;color:var(--pw-brand-muted);letter-spacing:-.01em">${escHtml(l)}</span>`).join("\n")}
 </div>`;
 }
+function renderAnimatedStatHtml(slots, t) {
+  const value = getSlotValue(slots, t, "value");
+  const label = getSlotValue(slots, t, "label");
+  const unit = getSlotValue(slots, t, "unit");
+  const sublabel = getSlotValue(slots, t, "sublabel");
+  const dir = getSlotValue(slots, t, "delta_direction");
+  const delta = getSlotValue(slots, t, "delta");
+  const arrow = dir === "up" ? "\u2191" : dir === "down" ? "\u2193" : "";
+  const deltaColor = dir === "up" ? "#22c55e" : dir === "down" ? "#ef4444" : "var(--pw-brand-muted)";
+  return `<div style="display:flex;flex-direction:column;align-items:center;justify-content:center;min-height:240px;text-align:center;gap:var(--pw-spacing-sm)">
+  <div id="pw-anim-value" style="font-size:5rem;font-weight:900;line-height:1;font-family:var(--pw-font-heading);color:var(--pw-brand-primary);letter-spacing:-0.04em">0</div>
+  ${unit ? `<div style="font-size:1.25rem;font-weight:600;color:var(--pw-brand-muted);margin-top:-var(--pw-spacing-sm)">${escHtml(unit)}</div>` : ""}
+  <div style="font-size:1.1rem;font-weight:600;color:var(--pw-brand-text);letter-spacing:0.01em">${escHtml(label)}</div>
+  ${sublabel ? `<div style="font-size:.8rem;color:var(--pw-brand-muted)">${escHtml(sublabel)}</div>` : ""}
+  ${delta ? `<div style="font-size:.875rem;color:${deltaColor};font-weight:600">${arrow}${arrow ? " " : ""}${escHtml(delta)}</div>` : ""}
+</div>
+<script>
+(function(){
+  var target = ${Number(value.replace(/[^0-9.-]/g, "")) || 0};
+  var isFloat = target % 1 !== 0;
+  var obj = { val: 0 };
+  anime({
+    targets: obj,
+    val: target,
+    duration: ${2e3},
+    easing: "easeOutExpo",
+    update: function() {
+      document.getElementById("pw-anim-value").textContent =
+        isFloat ? obj.val.toFixed(1) : Math.round(obj.val).toLocaleString();
+    }
+  });
+})();
+</script>`;
+}
+function renderReleaseCardHtml(slots, t) {
+  const version = getSlotValue(slots, t, "version");
+  const headline = getSlotValue(slots, t, "headline");
+  const changes = getSlotValue(slots, t, "changes").split("\n").filter(Boolean);
+  const tag = getSlotValue(slots, t, "tag");
+  const tagColors = { feat: "#22c55e", fix: "#f59e0b", perf: "#60a5fa", break: "#ef4444" };
+  const tc = tagColors[tag] ?? "var(--pw-brand-accent)";
+  return `<div class="pw-release-wrap" style="display:flex;flex-direction:column;gap:var(--pw-spacing-md)">
+  <div style="display:flex;align-items:center;gap:var(--pw-spacing-sm);opacity:0" class="pw-ri">
+    <span style="font-family:var(--pw-font-mono);font-size:1.1rem;font-weight:700;color:var(--pw-brand-primary)">${escHtml(version)}</span>
+    ${tag ? `<span style="font-size:.7rem;padding:2px 8px;border-radius:var(--pw-radius-full);background:${tc}22;color:${tc};font-weight:700;text-transform:uppercase;letter-spacing:.06em">${escHtml(tag)}</span>` : ""}
+  </div>
+  <h2 style="margin:0;font-size:1.5rem;font-weight:800;line-height:1.25;font-family:var(--pw-font-heading);opacity:0" class="pw-ri">${escHtml(headline)}</h2>
+  <ul style="margin:0;padding:0;list-style:none;display:flex;flex-direction:column;gap:var(--pw-spacing-xs)">
+    ${changes.map((c) => `<li class="pw-change-item" style="display:flex;align-items:flex-start;gap:var(--pw-spacing-xs);font-size:.9rem;opacity:0"><span style="color:${tc};flex-shrink:0;margin-top:2px">\u2192</span><span>${escHtml(c)}</span></li>`).join("\n    ")}
+  </ul>
+</div>
+<script>
+(function(){
+  anime({ targets: ".pw-ri", translateY: [-16, 0], opacity: [0, 1], delay: anime.stagger(120), duration: 500, easing: "easeOutCubic" });
+  anime({ targets: ".pw-change-item", translateX: [-20, 0], opacity: [0, 1], delay: anime.stagger(80, { start: 350 }), duration: 450, easing: "easeOutCubic" });
+})();
+</script>`;
+}
+function renderChartBarsHtml(slots, t) {
+  const title = getSlotValue(slots, t, "title");
+  const labelsRaw = getSlotValue(slots, t, "labels").split("\n").filter(Boolean);
+  const valuesRaw = getSlotValue(slots, t, "values").split("\n").filter(Boolean).map(Number);
+  const orientation = getSlotValue(slots, t, "orientation") || "vertical";
+  getSlotValue(slots, t, "accent_color") || "var(--pw-brand-primary)";
+  const labelsJson = JSON.stringify(labelsRaw);
+  const valuesJson = JSON.stringify(valuesRaw);
+  const chartType = orientation === "horizontal" ? "bar" : "bar";
+  const indexAxis = orientation === "horizontal" ? `indexAxis: "y",` : "";
+  return `${title ? `<h3 style="margin:0 0 var(--pw-spacing-md);font-size:1rem;font-weight:700;font-family:var(--pw-font-heading);color:var(--pw-brand-text)">${escHtml(title)}</h3>` : ""}
+<div style="position:relative;width:100%;height:220px">
+  <canvas id="pw-chart" style="width:100%;height:100%"></canvas>
+</div>
+<script>
+(function(){
+  var ctx = document.getElementById("pw-chart").getContext("2d");
+  var accent = getComputedStyle(document.querySelector(".pw-widget")).getPropertyValue("--pw-brand-primary").trim() || "#f97316";
+  new Chart(ctx, {
+    type: "${chartType}",
+    data: {
+      labels: ${labelsJson},
+      datasets: [{
+        data: ${valuesJson},
+        backgroundColor: accent + "cc",
+        borderColor: accent,
+        borderWidth: 2,
+        borderRadius: 6,
+      }]
+    },
+    options: {
+      ${indexAxis}
+      responsive: true,
+      maintainAspectRatio: false,
+      animation: { duration: 1500, easing: "easeOutQuart" },
+      plugins: { legend: { display: false } },
+      scales: {
+        x: { grid: { color: "rgba(255,255,255,.06)" }, ticks: { color: "rgba(255,255,255,.5)", font: { size: 11 } } },
+        y: { grid: { color: "rgba(255,255,255,.06)" }, ticks: { color: "rgba(255,255,255,.5)", font: { size: 11 } } }
+      }
+    }
+  });
+})();
+</script>`;
+}
+var LUCIDE_ICON_MAP = {
+  typescript: "code-2",
+  javascript: "code-2",
+  "js": "code-2",
+  "ts": "code-2",
+  react: "atom",
+  next: "triangle",
+  nextjs: "triangle",
+  "next.js": "triangle",
+  node: "server",
+  nodejs: "server",
+  "node.js": "server",
+  postgres: "database",
+  postgresql: "database",
+  mysql: "database",
+  sqlite: "database",
+  supabase: "database",
+  prisma: "layers",
+  tailwind: "palette",
+  css: "palette",
+  sass: "palette",
+  docker: "box",
+  kubernetes: "cloud",
+  aws: "cloud",
+  vercel: "triangle",
+  github: "git-branch",
+  git: "git-branch",
+  graphql: "share-2",
+  rest: "globe",
+  api: "globe",
+  redis: "zap",
+  kafka: "radio",
+  python: "terminal",
+  rust: "cpu",
+  go: "activity",
+  java: "coffee",
+  vue: "layers",
+  svelte: "layers",
+  angular: "layers",
+  figma: "pen-tool",
+  linear: "layers",
+  default: "layers"
+};
+function renderArchitectureBadgeHtml(slots, t) {
+  const title = getSlotValue(slots, t, "title");
+  const subtitle = getSlotValue(slots, t, "subtitle");
+  const items = getSlotValue(slots, t, "items").split("\n").filter(Boolean);
+  const palette = ["#60a5fa", "#f97316", "#34d399", "#a78bfa", "#fb923c", "#38bdf8", "#facc15", "#f472b6"];
+  const chips = items.map((item, i) => {
+    const key = item.trim().toLowerCase().replace(/[^a-z0-9.]/g, "");
+    const icon = LUCIDE_ICON_MAP[key] ?? LUCIDE_ICON_MAP.default;
+    const color = palette[i % palette.length];
+    return `<div class="pw-arch-item" style="display:inline-flex;align-items:center;gap:6px;padding:6px 14px;border-radius:var(--pw-radius-full);border:1px solid ${color}33;background:${color}11;opacity:0">
+      <i data-lucide="${icon}" style="width:14px;height:14px;color:${color};flex-shrink:0"></i>
+      <span style="font-size:.8rem;font-weight:600;color:${color};letter-spacing:.02em">${escHtml(item.trim())}</span>
+    </div>`;
+  }).join("\n    ");
+  return `<div style="display:flex;flex-direction:column;gap:var(--pw-spacing-md)">
+  ${title ? `<h3 style="margin:0;font-size:1.1rem;font-weight:700;font-family:var(--pw-font-heading);color:var(--pw-brand-text);opacity:0" id="pw-arch-title">${escHtml(title)}</h3>` : ""}
+  <div style="display:flex;flex-wrap:wrap;gap:var(--pw-spacing-xs)">
+    ${chips}
+  </div>
+  ${subtitle ? `<p style="margin:0;font-size:.75rem;color:var(--pw-brand-muted);opacity:0" class="pw-arch-sub">${escHtml(subtitle)}</p>` : ""}
+</div>
+<script>
+(function(){
+  lucide.createIcons();
+  var title = document.getElementById("pw-arch-title");
+  if(title) anime({ targets: title, opacity: [0,1], translateY: [-10,0], duration: 400, easing: "easeOutCubic" });
+  anime({ targets: ".pw-arch-item", opacity: [0,1], scale: [0.85,1], delay: anime.stagger(70, { start: 200 }), duration: 400, easing: "easeOutBack" });
+  var sub = document.querySelector(".pw-arch-sub");
+  if(sub) anime({ targets: sub, opacity: [0,1], delay: 600, duration: 400, easing: "easeOutCubic" });
+})();
+</script>`;
+}
 var HTML_BODY_FN = {
   "stat-card": renderStatCardHtml,
   "feature-grid": renderFeatureGridHtml,
@@ -2080,7 +4759,11 @@ var HTML_BODY_FN = {
   "metric-badge": renderMetricBadgeHtml,
   "pricing-tier": renderPricingTierHtml,
   "changelog-row": renderChangelogRowHtml,
-  "social-proof": renderSocialProofHtml
+  "social-proof": renderSocialProofHtml,
+  "animated-stat": renderAnimatedStatHtml,
+  "release-card": renderReleaseCardHtml,
+  "chart-bars": renderChartBarsHtml,
+  "architecture-badge": renderArchitectureBadgeHtml
 };
 function renderStatCardMd(slots, t) {
   const value = getSlotValue(slots, t, "value");
@@ -2151,6 +4834,43 @@ function renderSocialProofMd(slots, t) {
   const logos = getSlotValue(slots, t, "logos").split("\n").filter(Boolean);
   return `${caption ? caption + "\n\n" : ""}${logos.join(" \xB7 ")}`;
 }
+function renderAnimatedStatMd(slots, t) {
+  const value = getSlotValue(slots, t, "value");
+  const label = getSlotValue(slots, t, "label");
+  const unit = getSlotValue(slots, t, "unit");
+  const delta = getSlotValue(slots, t, "delta");
+  const dir = getSlotValue(slots, t, "delta_direction");
+  const arrow = dir === "up" ? "\u2191" : dir === "down" ? "\u2193" : "";
+  return `**${value}${unit ? " " + unit : ""}** \u2014 ${label}${delta ? ` ${arrow} ${delta}` : ""}`;
+}
+function renderReleaseCardMd(slots, t) {
+  const version = getSlotValue(slots, t, "version");
+  const headline = getSlotValue(slots, t, "headline");
+  const changes = getSlotValue(slots, t, "changes").split("\n").filter(Boolean);
+  const tag = getSlotValue(slots, t, "tag");
+  return `### ${version}${tag ? ` \`${tag}\`` : ""} \u2014 ${headline}
+
+${changes.map((c) => `- ${c}`).join("\n")}`;
+}
+function renderChartBarsMd(slots, t) {
+  const title = getSlotValue(slots, t, "title");
+  const labels = getSlotValue(slots, t, "labels").split("\n").filter(Boolean);
+  const values = getSlotValue(slots, t, "values").split("\n").filter(Boolean);
+  const rows = labels.map((l, i) => `- **${l}**: ${values[i] ?? "0"}`).join("\n");
+  return `${title ? `## ${title}
+
+` : ""}${rows}`;
+}
+function renderArchitectureBadgeMd(slots, t) {
+  const title = getSlotValue(slots, t, "title");
+  const items = getSlotValue(slots, t, "items").split("\n").filter(Boolean);
+  const subtitle = getSlotValue(slots, t, "subtitle");
+  return `${title ? `**${title}**
+
+` : ""}${items.map((i) => `- ${i.trim()}`).join("\n")}${subtitle ? `
+
+*${subtitle}*` : ""}`;
+}
 var MD_FN = {
   "stat-card": renderStatCardMd,
   "feature-grid": renderFeatureGridMd,
@@ -2159,7 +4879,11 @@ var MD_FN = {
   "metric-badge": renderMetricBadgeMd,
   "pricing-tier": renderPricingTierMd,
   "changelog-row": renderChangelogRowMd,
-  "social-proof": renderSocialProofMd
+  "social-proof": renderSocialProofMd,
+  "animated-stat": renderAnimatedStatMd,
+  "release-card": renderReleaseCardMd,
+  "chart-bars": renderChartBarsMd,
+  "architecture-badge": renderArchitectureBadgeMd
 };
 function toReactStyle(vars) {
   return "{\n" + Object.entries(vars).map(([k, v]) => `    "${k}": "${v.replace(/"/g, '\\"')}"`).join(",\n") + "\n  }";
@@ -2191,6 +4915,89 @@ ${slotDecls}
 }
 `;
 }
+function getRequiredLibs(templateId) {
+  const libs = /* @__PURE__ */ new Set(["tailwind"]);
+  if (["animated-stat", "release-card"].includes(templateId)) {
+    libs.add("anime");
+  }
+  if (templateId === "chart-bars") {
+    libs.add("anime");
+    libs.add("chartjs");
+  }
+  if (templateId === "architecture-badge") {
+    libs.add("anime");
+    libs.add("lucide");
+  }
+  return Array.from(libs);
+}
+function buildCdnScripts(libs) {
+  const tags = [];
+  if (libs.includes("tailwind")) {
+    tags.push(`<script>window.tailwind={config:{},silent:true}</script>`);
+    tags.push(`<script src="https://cdn.tailwindcss.com"></script>`);
+  }
+  if (libs.includes("chartjs")) {
+    tags.push(`<script src="https://cdn.jsdelivr.net/npm/chart.js@4.4.2/dist/chart.umd.min.js"></script>`);
+  }
+  if (libs.includes("anime")) {
+    tags.push(`<script src="https://cdn.jsdelivr.net/npm/animejs@3.2.2/lib/anime.min.js"></script>`);
+  }
+  if (libs.includes("lucide")) {
+    tags.push(`<script src="https://unpkg.com/lucide@latest/dist/umd/lucide.min.js"></script>`);
+  }
+  if (libs.includes("alpine")) {
+    tags.push(`<script defer src="https://cdn.jsdelivr.net/npm/alpinejs@3.x.x/dist/cdn.min.js"></script>`);
+  }
+  return tags.join("\n  ");
+}
+function renderStandalone(template, slots, brand, style) {
+  const bodyFn = HTML_BODY_FN[template.id];
+  const body = bodyFn ? bodyFn(slots, template) : "";
+  const allVars = { ...brandVars(brand), ...styleVars(style) };
+  const inlineStyle = buildInlineStyle(allVars);
+  const customCss = style.customCss ?? "";
+  const override = style.templateOverrides?.[template.id]?.html;
+  const finalBody = override ?? body;
+  const libs = getRequiredLibs(template.id);
+  const scripts = buildCdnScripts(libs);
+  const w = template.defaultWidth ?? 600;
+  const h = template.defaultHeight ?? 400;
+  return `<!DOCTYPE html>
+<html lang="en">
+<head>
+  <meta charset="UTF-8" />
+  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+  <title>${escHtml(template.name)} \u2014 forge0x2B</title>
+  ${scripts}
+  <style>
+    *, *::before, *::after { box-sizing: border-box; }
+    html, body {
+      margin: 0; padding: 0;
+      width: ${w}px; height: ${h}px;
+      overflow: hidden;
+      background: transparent;
+      display: flex; align-items: center; justify-content: center;
+    }
+    .pw-widget {
+      font-family: var(--pw-font-body);
+      color: var(--pw-brand-text);
+      background: var(--pw-brand-surface);
+      border-radius: var(--pw-radius-md);
+      padding: var(--pw-spacing-md);
+      border: var(--pw-border-width) solid rgba(255,255,255,.08);
+      box-sizing: border-box;
+      width: 100%; max-width: ${w}px;
+      ${customCss}
+    }
+  </style>
+</head>
+<body>
+  <div class="pw-widget pw-${template.id}" style="${inlineStyle}">
+    ${finalBody}
+  </div>
+</body>
+</html>`;
+}
 function renderWidget(input) {
   const { template, slots, brandKit, format } = input;
   const style = !input.style ? STYLE_PRESET_DEFAULT : typeof input.style === "string" ? getStyleById(input.style) : input.style;
@@ -2201,6 +5008,9 @@ function renderWidget(input) {
   if (format === "react") {
     return renderReact(template, slots, brandKit ?? null, style);
   }
+  if (format === "standalone") {
+    return renderStandalone(template, slots, brandKit ?? null, style);
+  }
   const bodyFn = HTML_BODY_FN[template.id];
   const body = bodyFn ? bodyFn(slots, template) : "";
   const allVars = { ...brandVars(brandKit ?? null), ...styleVars(style) };
@@ -2556,7 +5366,7 @@ function getDispatchChannel(id) {
 }
 
 // src/forge/dispatch.ts
-function buildSystemPrompt7(channel, audience, brand, blueprintContext, toneOffset) {
+function buildSystemPrompt14(channel, audience, brand, blueprintContext, toneOffset) {
   const lines = [
     `You are a skilled content writer producing a ${channel.name} post.`,
     "",
@@ -2607,7 +5417,7 @@ function truncate(content, maxLength) {
 async function generateForChannel(ask, channel, provider, opts) {
   const now = (/* @__PURE__ */ new Date()).toISOString();
   try {
-    const system = buildSystemPrompt7(
+    const system = buildSystemPrompt14(
       channel,
       opts.audience,
       opts.brand,
@@ -2752,7 +5562,7 @@ function nextVersionNumber(versions) {
 }
 
 // src/forge/refineAsset.ts
-function buildSystemPrompt8(input) {
+function buildSystemPrompt15(input) {
   const parts = [
     `You are a content refinement assistant for a developer-focused content tool (forge0x2B).`,
     `You are refining an existing ${input.assetType.replace(/-/g, " ")} asset.`,
@@ -2791,7 +5601,7 @@ function parseRefinedResponse(raw) {
   return { newContent, llmReply };
 }
 async function refineAsset(input, provider) {
-  const systemPrompt = buildSystemPrompt8(input);
+  const systemPrompt = buildSystemPrompt15(input);
   if (scanForSecrets(systemPrompt)) {
     throw new Error("refineAsset: secret pattern detected in asset content. Refusing to send to LLM.");
   }
@@ -2953,4 +5763,4 @@ function previewExport(entries, format) {
   }
 }
 
-export { ANIMATION_DURATION_PRESETS, BRAND_CONTENT_SLOT_KEYS, BUNDLED_DISPATCH_CHANNELS, BUNDLED_STYLE_PRESETS, BUNDLED_WIDGET_TEMPLATES, DEFAULT_ANIMATION_DURATION_SECONDS, DEFAULT_BRAND_KIT_FONTS, DEFAULT_BRAND_KIT_PALETTE, DEFAULT_BRAND_KIT_VOICE, DISPATCH_CHANNEL_BLOG, DISPATCH_CHANNEL_EMAIL, DISPATCH_CHANNEL_HN, DISPATCH_CHANNEL_INSTAGRAM, DISPATCH_CHANNEL_LINKEDIN, DISPATCH_CHANNEL_NEWSLETTER, DISPATCH_CHANNEL_REDDIT, DISPATCH_CHANNEL_SLACK, DISPATCH_CHANNEL_TWEET, FORGE_AUDIENCES, FORGE_BRAND_THEME_ID, FORGE_TEMPLATES, FREE_TIER_WIDGET_IDS, MAX_ANIMATION_DURATION_SECONDS, MIN_ANIMATION_DURATION_SECONDS, PRISM_TEMPLATES, PRISM_TEMPLATE_ARCHITECTURE_OVERVIEW, PRISM_TEMPLATE_REFACTOR_RATIONALE, PRISM_TEMPLATE_RELEASE_ANNOUNCEMENT, PRISM_TEMPLATE_SHIPPING_DIGEST, PRISM_TEMPLATE_ZONE_DEEPDIVE, REFINE_COUNTDOWN_THRESHOLD, REFINE_SESSION_SOFT_CAP, REFINE_SUGGESTIONS, STYLE_PRESET_BRUTALIST, STYLE_PRESET_DEFAULT, STYLE_PRESET_GLASSY, STYLE_PRESET_MINIMAL, TEMPLATE_SCHEMA_EXAMPLES, TEMPLATE_SCHEMA_GENERIC, WIDGET_TEMPLATE_CHANGELOG_ROW, WIDGET_TEMPLATE_CTA_BANNER, WIDGET_TEMPLATE_FEATURE_GRID, WIDGET_TEMPLATE_METRIC_BADGE, WIDGET_TEMPLATE_PRICING_TIER, WIDGET_TEMPLATE_SOCIAL_PROOF, WIDGET_TEMPLATE_STAT_CARD, WIDGET_TEMPLATE_TESTIMONIAL, applyEntryPatch, asAudienceId, assembleBrandUrlExtractionPrompt, assembleForgePrompt, brandThemeConfigToEntry, buildRevertVersion, buildScheduledEntry, buildVersion, cascadingScheduledFor, clampAnimationDuration, computeDiff, defaultBrandKit, defaultValueForField, distill, entryInRange, exportToBufferCsv, exportToHypefuryCsv, exportToICalendar, generateArchitectureWalkthrough, generateAskDrivenAsset, generateChangesSince, generateOnboardingDoc, generateRefactoringReport, generateReleaseNotes, getDispatchChannel, getPrismTemplate, getSlotValue, getStyleById, getWidgetTemplate, initialFormValues, next7DaysRange, nextVersionNumber, orchestrateDispatch, parseAndValidateSchemaDefinition, parseBrandKitFromLlmResponse, parseBrandThemeContent, parseStyleFromCss, parseStyleFromTailwindConfig, parseStyleFromTokensJson, parseThemeConfigContent, previewExport, refineAsset, refineLimitState, renderWidget, resolveAnimatedChoice, resolveAnimationDuration, resolveBrandPalette, runForgeGeneration, scanForSecrets, schemaExampleFor, schemaToForm, templateAnimatedDefault, themeEntryToPalette, tryParseJsonObject, validateAgainstTemplateSchema, validateAssetSlots, validateFormValues, validateSchemaDefinition };
+export { ANIMATION_DURATION_PRESETS, BRAND_CONTENT_SLOT_KEYS, BUNDLED_DISPATCH_CHANNELS, BUNDLED_STYLE_PRESETS, BUNDLED_WIDGET_TEMPLATES, DEFAULT_ANIMATION_DURATION_SECONDS, DEFAULT_BRAND_KIT_FONTS, DEFAULT_BRAND_KIT_PALETTE, DEFAULT_BRAND_KIT_VOICE, DISPATCH_CHANNEL_BLOG, DISPATCH_CHANNEL_EMAIL, DISPATCH_CHANNEL_HN, DISPATCH_CHANNEL_INSTAGRAM, DISPATCH_CHANNEL_LINKEDIN, DISPATCH_CHANNEL_NEWSLETTER, DISPATCH_CHANNEL_REDDIT, DISPATCH_CHANNEL_SLACK, DISPATCH_CHANNEL_TWEET, FORGE_AUDIENCES, FORGE_BRAND_THEME_ID, FORGE_TEMPLATES, FREE_TIER_WIDGET_IDS, MAX_ANIMATION_DURATION_SECONDS, MIN_ANIMATION_DURATION_SECONDS, PRISM_TEMPLATES, PRISM_TEMPLATE_ARCHITECTURE_OVERVIEW, PRISM_TEMPLATE_REFACTOR_RATIONALE, PRISM_TEMPLATE_RELEASE_ANNOUNCEMENT, PRISM_TEMPLATE_SHIPPING_DIGEST, PRISM_TEMPLATE_ZONE_DEEPDIVE, REFINE_COUNTDOWN_THRESHOLD, REFINE_SESSION_SOFT_CAP, REFINE_SUGGESTIONS, STYLE_PRESET_BRUTALIST, STYLE_PRESET_DEFAULT, STYLE_PRESET_GLASSY, STYLE_PRESET_MINIMAL, TEMPLATE_SCHEMA_EXAMPLES, TEMPLATE_SCHEMA_GENERIC, WIDGET_TEMPLATE_CHANGELOG_ROW, WIDGET_TEMPLATE_CTA_BANNER, WIDGET_TEMPLATE_FEATURE_GRID, WIDGET_TEMPLATE_METRIC_BADGE, WIDGET_TEMPLATE_PRICING_TIER, WIDGET_TEMPLATE_SOCIAL_PROOF, WIDGET_TEMPLATE_STAT_CARD, WIDGET_TEMPLATE_TESTIMONIAL, applyEntryPatch, asAudienceId, assembleBrandUrlExtractionPrompt, assembleForgePrompt, brandThemeConfigToEntry, buildRevertVersion, buildScheduledEntry, buildVersion, cascadingScheduledFor, clampAnimationDuration, computeDiff, defaultBrandKit, defaultValueForField, distill, entryInRange, exportToBufferCsv, exportToHypefuryCsv, exportToICalendar, generateADR, generateArc42, generateArchitectureWalkthrough, generateAskDrivenAsset, generateChangesSince, generateComplianceDoc, generateKnowledgeCapture, generateNewsletter, generateOnboardingDoc, generatePresentation, generateRadio, generateRefactoringReport, generateReleaseNotes, generateSprintRetro, getDispatchChannel, getPrismTemplate, getSlotValue, getStyleById, getWidgetTemplate, initialFormValues, next7DaysRange, nextVersionNumber, orchestrateDispatch, parseAndValidateSchemaDefinition, parseBrandKitFromLlmResponse, parseBrandThemeContent, parseStyleFromCss, parseStyleFromTailwindConfig, parseStyleFromTokensJson, parseThemeConfigContent, previewExport, refineAsset, refineLimitState, renderWidget, resolveAnimatedChoice, resolveAnimationDuration, resolveBrandPalette, runForgeGeneration, scanForSecrets, schemaExampleFor, schemaToForm, templateAnimatedDefault, themeEntryToPalette, tryParseJsonObject, validateAgainstTemplateSchema, validateAssetSlots, validateFormValues, validateSchemaDefinition };