ai-spec-dev 0.46.0 → 0.55.0

package/core/import-verifier.ts

@@ -0,0 +1,569 @@
+import * as fs from "fs-extra";
+import * as path from "path";
+import chalk from "chalk";
+
+// ─── Types ────────────────────────────────────────────────────────────────────
+
+export interface ImportRef {
+  /** Raw module specifier as written in source: '@/apis/foo' */
+  source: string;
+  /** Absolute path to the resolved file (when resolution succeeded) */
+  resolvedPath?: string;
+  /** Names imported from this module (empty for side-effect or default-only) */
+  importedNames: string[];
+  /** True for `import type { ... }` */
+  isTypeOnly: boolean;
+  /** True when default import is present: `import X from '...'` */
+  hasDefault: boolean;
+  /** Default import local name (when hasDefault is true) */
+  defaultName?: string;
+  /** File where this import is declared (relative to repo root) */
+  file: string;
+  /** 1-indexed line number */
+  line: number;
+}
+
+export interface BrokenImport {
+  ref: ImportRef;
+  reason: "file_not_found" | "missing_export";
+  /** When reason === "missing_export": which named imports are missing */
+  missingExports?: string[];
+  /**
+   * When reason === "missing_export": the full list of names the target file
+   * DOES export. Used by import-fixer to detect rename-style fixes (e.g.
+   * import `{ Task }` but target exports `{ TaskItem }`).
+   */
+  availableExports?: string[];
+  /** Suggestion for what file/path the AI may have intended */
+  suggestion?: string;
+}
+
+export interface ImportVerificationReport {
+  totalFiles: number;
+  totalImports: number;
+  /** Imports skipped because the source is an external package */
+  externalImports: number;
+  /** Imports successfully resolved + (when applicable) named exports validated */
+  matchedImports: number;
+  brokenImports: BrokenImport[];
+}
+
+// ─── tsconfig path alias resolution ───────────────────────────────────────────
+
+interface PathAliases {
+  baseUrl: string;
+  /** Map: alias prefix (with trailing /*) → target prefix */
+  paths: Array<{ alias: string; target: string }>;
+}
+
+/**
+ * Strip JSON-with-comments to plain JSON. Handles // line comments,
+ * /* block comments, and trailing commas.
+ */
+function stripJsonComments(src: string): string {
+  // remove /* ... */ comments
+  let out = src.replace(/\/\*[\s\S]*?\*\//g, "");
+  // remove // ... line comments
+  out = out.replace(/(^|[^:])\/\/[^\n]*/g, "$1");
+  // remove trailing commas before } or ]
+  out = out.replace(/,(\s*[}\]])/g, "$1");
+  return out;
+}
+
+/**
+ * Load path aliases from tsconfig.json / tsconfig.app.json / jsconfig.json.
+ * Falls back to a sensible default mapping `@/*` → `src/*` if no config found.
+ */
+export async function loadPathAliases(repoRoot: string): Promise<PathAliases> {
+  const candidates = ["tsconfig.json", "tsconfig.app.json", "jsconfig.json"];
+  for (const name of candidates) {
+    const p = path.join(repoRoot, name);
+    if (!(await fs.pathExists(p))) continue;
+    try {
+      const raw = await fs.readFile(p, "utf-8");
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      const cfg: any = JSON.parse(stripJsonComments(raw));
+      const baseUrl = cfg?.compilerOptions?.baseUrl ?? ".";
+      const paths = cfg?.compilerOptions?.paths ?? {};
+      const entries: Array<{ alias: string; target: string }> = [];
+      for (const [aliasKey, targets] of Object.entries(paths)) {
+        if (!Array.isArray(targets) || targets.length === 0) continue;
+        const target = String((targets as string[])[0]);
+        entries.push({ alias: aliasKey, target });
+      }
+      if (entries.length > 0) {
+        return { baseUrl, paths: entries };
+      }
+    } catch {
+      // ignore parse errors, try next file
+    }
+  }
+
+  // Default fallback: most Vue/React projects use `@/*` → `src/*`
+  return {
+    baseUrl: ".",
+    paths: [{ alias: "@/*", target: "src/*" }],
+  };
+}
+
+/**
+ * Resolve an import specifier to an absolute candidate path (without extension).
+ * Returns null if the specifier is an external package.
+ */
+export function resolveSpecifier(
+  specifier: string,
+  fromFileAbs: string,
+  repoRoot: string,
+  aliases: PathAliases
+): string | null {
+  // Relative import
+  if (specifier.startsWith("./") || specifier.startsWith("../")) {
+    return path.resolve(path.dirname(fromFileAbs), specifier);
+  }
+
+  // Absolute path (rare in source code, but handle it)
+  if (specifier.startsWith("/")) {
+    return specifier;
+  }
+
+  // Alias-based import
+  for (const { alias, target } of aliases.paths) {
+    const aliasPrefix = alias.replace(/\*$/, "");
+    if (specifier.startsWith(aliasPrefix)) {
+      const remainder = specifier.slice(aliasPrefix.length);
+      const targetPrefix = target.replace(/\*$/, "");
+      const baseAbs = path.resolve(repoRoot, aliases.baseUrl);
+      return path.resolve(baseAbs, targetPrefix + remainder);
+    }
+  }
+
+  // External package — skip
+  return null;
+}
+
+/**
+ * Try to resolve a candidate path to an actual file by trying common extensions
+ * and index files (Node-style resolution).
+ */
+const CANDIDATE_EXTENSIONS = [
+  "",
+  ".ts", ".tsx", ".js", ".jsx", ".vue", ".mjs", ".mts", ".d.ts",
+];
+const INDEX_NAMES = ["index.ts", "index.tsx", "index.js", "index.jsx", "index.vue", "index.mjs"];
+
+export async function resolveToActualFile(candidate: string): Promise<string | null> {
+  // 1. Try the path as-is with each extension
+  for (const ext of CANDIDATE_EXTENSIONS) {
+    const p = candidate + ext;
+    try {
+      const stat = await fs.stat(p);
+      if (stat.isFile()) return p;
+    } catch { /* not found */ }
+  }
+
+  // 2. Try as a directory + index file
+  try {
+    const stat = await fs.stat(candidate);
+    if (stat.isDirectory()) {
+      for (const name of INDEX_NAMES) {
+        const p = path.join(candidate, name);
+        try {
+          if ((await fs.stat(p)).isFile()) return p;
+        } catch { /* not found */ }
+      }
+    }
+  } catch { /* not a directory */ }
+
+  return null;
+}
+
+// ─── Import statement parsing ─────────────────────────────────────────────────
+
+/**
+ * Extract `<script>` and `<script setup>` block contents from a Vue SFC.
+ * Returns an array of [startLine, content] pairs so line numbers stay aligned
+ * with the original file.
+ */
+function extractVueScriptBlocks(source: string): Array<{ startLine: number; content: string }> {
+  const blocks: Array<{ startLine: number; content: string }> = [];
+  const re = /<script[^>]*>([\s\S]*?)<\/script>/g;
+  let m: RegExpExecArray | null;
+  while ((m = re.exec(source)) !== null) {
+    const before = source.slice(0, m.index);
+    const startLine = before.split("\n").length;
+    // Skip the opening <script ...> tag itself (find first newline after match start)
+    const tagEnd = source.indexOf(">", m.index) + 1;
+    const contentLine = source.slice(0, tagEnd).split("\n").length;
+    blocks.push({ startLine: contentLine, content: m[1] });
+  }
+  return blocks;
+}
+
+/**
+ * Parse all `import ... from '...'` statements in a JS/TS source.
+ * Returns ImportRefs with line numbers (1-indexed) relative to `source`.
+ */
+export function parseImports(source: string, fileRel: string): ImportRef[] {
+  const refs: ImportRef[] = [];
+  const lines = source.split("\n");
+
+  // Walk line-by-line for accurate line numbers (handles multi-line imports too)
+  // by joining continuation lines until we close the import.
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i];
+    const trimmed = line.trim();
+    if (!trimmed.startsWith("import ") && trimmed !== "import" && !trimmed.startsWith("import{") && !trimmed.startsWith("import(")) {
+      continue;
+    }
+
+    // Greedily collect lines until we find the matching `from '...'` or `from "..."` or end-of-statement
+    let block = line;
+    let j = i;
+    while (j < lines.length - 1 && !/from\s+['"`]/.test(block) && !/^\s*import\s+['"`]/.test(block)) {
+      j++;
+      block += "\n" + lines[j];
+      if (block.length > 2000) break; // safety
+    }
+
+    // Match: import ... from '...'
+    const fromMatch = block.match(
+      /^\s*import\s+(type\s+)?([^'"`]*?)\s+from\s+['"`]([^'"`]+)['"`]/
+    );
+    // Match: import '...' (side-effect)
+    const sideEffectMatch = block.match(/^\s*import\s+['"`]([^'"`]+)['"`]/);
+
+    if (fromMatch) {
+      const isTypeOnly = !!fromMatch[1];
+      const importClause = fromMatch[2].trim();
+      const sourceSpec = fromMatch[3];
+      const { defaultName, named } = parseImportClause(importClause);
+      refs.push({
+        source: sourceSpec,
+        importedNames: named,
+        isTypeOnly,
+        hasDefault: !!defaultName,
+        defaultName,
+        file: fileRel,
+        line: i + 1,
+      });
+    } else if (sideEffectMatch) {
+      refs.push({
+        source: sideEffectMatch[1],
+        importedNames: [],
+        isTypeOnly: false,
+        hasDefault: false,
+        file: fileRel,
+        line: i + 1,
+      });
+    }
+
+    i = j;
+  }
+
+  return refs;
+}
+
+/**
+ * Parse the part between `import` and `from`.
+ *
+ *  "X"                       → default X, named []
+ *  "{ A, B as C }"           → default undef, named [A, B]  (original names, not local bindings)
+ *  "X, { A, B }"             → default X, named [A, B]
+ *  "* as ns"                 → default undef, named []  (namespace, treat as default-like)
+ *
+ * IMPORTANT: For `{ A as B }`, we return `A` (the ORIGINAL exported name), not `B`
+ * (the local binding). This is because the verifier uses these names to validate
+ * against the target file's exports — and the target exports A, not B.
+ */
+function parseImportClause(clause: string): { defaultName?: string; named: string[] } {
+  const result: { defaultName?: string; named: string[] } = { named: [] };
+  if (!clause) return result;
+
+  // Strip namespace import — we treat it as opaque and don't validate names
+  if (/\*\s+as\s+\w+/.test(clause)) {
+    return result;
+  }
+
+  // Split on the first { to separate default from named
+  const bracePos = clause.indexOf("{");
+  let defaultPart = "";
+  let namedPart = "";
+  if (bracePos === -1) {
+    defaultPart = clause.trim();
+  } else {
+    // Strip trailing whitespace + comma (e.g. "React, " → "React")
+    defaultPart = clause.slice(0, bracePos).trim().replace(/,\s*$/, "").trim();
+    const closing = clause.indexOf("}", bracePos);
+    namedPart = clause.slice(bracePos + 1, closing === -1 ? undefined : closing);
+  }
+
+  if (defaultPart && /^\w+$/.test(defaultPart)) {
+    result.defaultName = defaultPart;
+  }
+
+  if (namedPart) {
+    const names = namedPart
+      .split(",")
+      .map((s) => s.trim())
+      .filter(Boolean)
+      .map((entry) => {
+        // "A as B" → use A (the ORIGINAL exported name, for export validation)
+        const asMatch = entry.match(/^(?:type\s+)?(\w+)\s+as\s+(\w+)$/);
+        if (asMatch) return asMatch[1];
+        // "type A" → A (drop the type modifier)
+        return entry.replace(/^type\s+/, "").trim();
+      })
+      .filter((n) => /^\w+$/.test(n));
+    result.named = names;
+  }
+
+  return result;
+}
+
+// ─── Export parsing (for named export validation) ─────────────────────────────
+
+/**
+ * Extract all named exports from a JS/TS source file.
+ * Used to verify that imports reference real exports.
+ *
+ *  export const X         → X
+ *  export function X      → X
+ *  export class X         → X
+ *  export interface X     → X
+ *  export type X          → X
+ *  export enum X          → X
+ *  export { A, B as C }   → A, C
+ *  export * from 'foo'    → __star__ (treated as wildcard, accepts anything)
+ *  export default ...     → default
+ */
+export function parseNamedExports(source: string): { names: Set<string>; hasWildcard: boolean; hasDefault: boolean } {
+  const names = new Set<string>();
+  let hasWildcard = false;
+  let hasDefault = false;
+
+  // export const|let|var|function|class|interface|type|enum NAME
+  const declRe = /\bexport\s+(?:async\s+)?(?:const|let|var|function\*?|class|interface|type|enum)\s+(\w+)/g;
+  let m: RegExpExecArray | null;
+  while ((m = declRe.exec(source)) !== null) {
+    names.add(m[1]);
+  }
+
+  // export { A, B as C, type D }
+  const blockRe = /\bexport\s*(?:type\s*)?\{([^}]*)\}/g;
+  while ((m = blockRe.exec(source)) !== null) {
+    const inner = m[1];
+    for (const part of inner.split(",")) {
+      const t = part.trim();
+      if (!t) continue;
+      const asMatch = t.match(/^(?:type\s+)?(\w+)\s+as\s+(\w+)$/);
+      if (asMatch) {
+        names.add(asMatch[2]);
+        continue;
+      }
+      const plain = t.replace(/^type\s+/, "").match(/^(\w+)/);
+      if (plain) names.add(plain[1]);
+    }
+  }
+
+  // export * from 'foo'
+  if (/\bexport\s*\*\s*from\s+['"`]/.test(source)) {
+    hasWildcard = true;
+  }
+
+  // export default ...
+  if (/\bexport\s+default\b/.test(source)) {
+    hasDefault = true;
+  }
+
+  return { names, hasWildcard, hasDefault };
+}
+
+// ─── Verification ─────────────────────────────────────────────────────────────
+
+/**
+ * Verify all imports in the given files actually resolve to existing files
+ * and reference real exports.
+ *
+ * @param files     Absolute paths of files to check (typically the freshly
+ *                  generated files from the codegen run).
+ * @param repoRoot  Absolute path to the repo root (used for tsconfig + alias).
+ */
+export async function verifyImports(
+  files: string[],
+  repoRoot: string
+): Promise<ImportVerificationReport> {
+  const aliases = await loadPathAliases(repoRoot);
+
+  let totalImports = 0;
+  let externalImports = 0;
+  let matchedImports = 0;
+  const broken: BrokenImport[] = [];
+  // Cache parsed exports per resolved file path
+  const exportsCache = new Map<string, ReturnType<typeof parseNamedExports>>();
+
+  // Build a set of generated file paths (resolved) so cross-file imports
+  // between fresh files can validate against each other even before they're
+  // physically written to disk in the same scan.
+  const generatedFileSet = new Set(files.map((f) => path.resolve(f)));
+
+  for (const fileAbs of files) {
+    let src: string;
+    try {
+      src = await fs.readFile(fileAbs, "utf-8");
+    } catch {
+      continue;
+    }
+    const fileRel = path.relative(repoRoot, fileAbs);
+
+    // For .vue files, only parse imports inside <script> blocks
+    let refs: ImportRef[];
+    if (fileAbs.endsWith(".vue")) {
+      refs = [];
+      for (const block of extractVueScriptBlocks(src)) {
+        const blockRefs = parseImports(block.content, fileRel);
+        // Adjust line numbers to match the original file
+        for (const r of blockRefs) {
+          r.line = block.startLine + r.line - 1;
+        }
+        refs.push(...blockRefs);
+      }
+    } else {
+      refs = parseImports(src, fileRel);
+    }
+
+    for (const ref of refs) {
+      totalImports++;
+
+      const candidate = resolveSpecifier(ref.source, fileAbs, repoRoot, aliases);
+      if (candidate === null) {
+        externalImports++;
+        continue;
+      }
+
+      const resolved = await resolveToActualFile(candidate);
+      if (!resolved) {
+        broken.push({
+          ref,
+          reason: "file_not_found",
+          suggestion: `expected at: ${path.relative(repoRoot, candidate)}.{ts,tsx,js,jsx,vue} or ${path.relative(repoRoot, candidate)}/index.*`,
+        });
+        continue;
+      }
+
+      ref.resolvedPath = resolved;
+
+      // Validate named exports (skip when no named imports were used)
+      if (ref.importedNames.length > 0) {
+        let exports = exportsCache.get(resolved);
+        if (!exports) {
+          try {
+            const targetSrc = await fs.readFile(resolved, "utf-8");
+            // For .vue targets, parse exports from <script> blocks too
+            const sourceForExports = resolved.endsWith(".vue")
+              ? extractVueScriptBlocks(targetSrc).map((b) => b.content).join("\n")
+              : targetSrc;
+            exports = parseNamedExports(sourceForExports);
+            exportsCache.set(resolved, exports);
+          } catch {
+            exports = { names: new Set(), hasWildcard: false, hasDefault: false };
+          }
+        }
+
+        // If the target re-exports from another module via `export *`, we can't
+        // be sure what's actually exported without recursive resolution.
+        // Treat wildcard exports as "trust the import" to avoid false positives.
+        if (!exports.hasWildcard) {
+          const missing = ref.importedNames.filter((n) => !exports!.names.has(n));
+          if (missing.length > 0) {
+            broken.push({
+              ref,
+              reason: "missing_export",
+              missingExports: missing,
+              availableExports: [...exports.names],
+              suggestion: exports.names.size > 0
+                ? `available exports: ${[...exports.names].slice(0, 8).join(", ")}${exports.names.size > 8 ? ", ..." : ""}`
+                : "target file has no named exports",
+            });
+            continue;
+          }
+        }
+      }
+
+      matchedImports++;
+    }
+  }
+
+  return {
+    totalFiles: files.length,
+    totalImports,
+    externalImports,
+    matchedImports,
+    brokenImports: broken,
+  };
+}
+
+// ─── Display ──────────────────────────────────────────────────────────────────
+
+export function printImportVerificationReport(
+  repoName: string,
+  report: ImportVerificationReport
+): void {
+  console.log(chalk.cyan(`\n─── Import Verification [${repoName}] ─────────────────────────`));
+  console.log(
+    chalk.gray(
+      `  Scanned ${report.totalFiles} generated file(s), checked ${report.totalImports} import(s)`
+    )
+  );
+  console.log(
+    chalk.gray(
+      `  External (skipped): ${report.externalImports}  ·  Internal verified: ${report.matchedImports}/${report.totalImports - report.externalImports}`
+    )
+  );
+
+  if (report.brokenImports.length === 0) {
+    console.log(chalk.green(`  ✔ All imports resolve correctly — 0 broken references.`));
+    console.log(chalk.cyan("─".repeat(65)));
+    return;
+  }
+
+  console.log(chalk.red(`\n  ❌ ${report.brokenImports.length} broken import(s):`));
+  const grouped = new Map<string, BrokenImport[]>();
+  for (const b of report.brokenImports) {
+    const key = b.ref.file;
+    if (!grouped.has(key)) grouped.set(key, []);
+    grouped.get(key)!.push(b);
+  }
+
+  let shownFiles = 0;
+  for (const [file, items] of grouped) {
+    if (shownFiles >= 10) {
+      console.log(chalk.gray(`     ... and ${grouped.size - shownFiles} more file(s) with broken imports`));
+      break;
+    }
+    console.log(chalk.yellow(`     ${file}`));
+    for (const b of items.slice(0, 4)) {
+      const reasonLabel = b.reason === "file_not_found" ? "file not found" : "missing export";
+      const namesLabel = b.reason === "missing_export"
+        ? ` { ${b.missingExports!.join(", ")} }`
+        : b.ref.importedNames.length > 0
+          ? ` { ${b.ref.importedNames.slice(0, 3).join(", ")}${b.ref.importedNames.length > 3 ? ", ..." : ""} }`
+          : "";
+      console.log(
+        chalk.gray(`       :${b.ref.line}  `) +
+        chalk.red(`${reasonLabel}`) +
+        chalk.gray(`${namesLabel} from '${b.ref.source}'`)
+      );
+      if (b.suggestion) {
+        console.log(chalk.gray(`              ↳ ${b.suggestion}`));
+      }
+    }
+    if (items.length > 4) {
+      console.log(chalk.gray(`       ... and ${items.length - 4} more in this file`));
+    }
+    shownFiles++;
+  }
+
+  console.log(chalk.gray(`\n  Tip: broken imports usually mean the AI hallucinated a file/export.`));
+  console.log(chalk.gray(`       Check whether the missing types/functions were declared inline elsewhere.`));
+  console.log(chalk.cyan("─".repeat(65)));
+}

package/core/knowledge-memory.ts

@@ -73,10 +73,14 @@ const MEMORY_SECTION_MARKER = "## 9. 积累教训";
  * Append review issues to the project constitution as accumulated lessons.
  * Creates the section if it doesn't exist; appends if it does.
  * Deduplicates by checking if a similar lesson already exists.
+ *
+ * @param reviewScore - The review score (0-10) at time of appending, embedded in each entry
+ *                      for future auditability ("was this lesson from a good or bad run?").
  */
 export async function appendLessonsToConstitution(
   projectRoot: string,
-  issues: ReviewIssue[]
+  issues: ReviewIssue[],
+  reviewScore?: number
 ): Promise<void> {
   if (issues.length === 0) return;
 
@@ -85,8 +89,37 @@ export async function appendLessonsToConstitution(
   try {
     content = await fs.readFile(constitutionPath, "utf-8");
   } catch {
-    console.log(chalk.gray("  No constitution file — skipping knowledge memory."));
-    return;
+    // No project-level constitution exists. This can happen when:
+    //   - ContextLoader reported "Constitution: found" based on a GLOBAL constitution
+    //     merged into the prompt, while the project itself has no file.
+    //   - The repo was never run through `ai-spec init`.
+    //
+    // Silently skipping accumulation means knowledge memory goes dark for this repo,
+    // so instead we create a minimal stub that subsequent runs can build on top of.
+    const stub = [
+      "# Project Constitution",
+      "",
+      "> Auto-generated stub. Run `ai-spec init` to populate §1–§8 with project-specific rules.",
+      "> §9 below is automatically accumulated from code review issues.",
+      "",
+      MEMORY_SECTION_HEADER.trim(),
+      "",
+    ].join("\n");
+    try {
+      await fs.writeFile(constitutionPath, stub, "utf-8");
+      content = stub;
+      console.log(
+        chalk.cyan(`  Created project constitution stub at ${constitutionPath}`)
+      );
+      console.log(
+        chalk.gray(`  (knowledge memory needs a project-level file to persist §9 lessons)`)
+      );
+    } catch (err) {
+      console.log(
+        chalk.yellow(`  ⚠ Could not create constitution stub: ${(err as Error).message}`)
+      );
+      return;
+    }
   }
 
   // Check if section 9 already exists
@@ -105,7 +138,8 @@ export async function appendLessonsToConstitution(
                   issue.category === "performance" ? "⚡" :
                   issue.category === "bug" ? "🐛" :
                   issue.category === "pattern" ? "📐" : "📝";
-    newEntries.push(`- ${badge} **[${date}]** ${issue.description}`);
+    const scoreTag = reviewScore !== undefined ? ` (r:${reviewScore.toFixed(1)})` : "";
+    newEntries.push(`- ${badge} **[${date}]**${scoreTag} ${issue.description}`);
   }
 
   if (newEntries.length === 0) {
@@ -203,6 +237,10 @@ export async function appendDirectLesson(
 
 /**
  * Full knowledge memory flow: extract issues from review → append to constitution.
+ *
+ * Quality gate: if the review score is ≥ 9.0, the run was excellent and the
+ * extracted issues are likely minor style nits. Skip accumulation to prevent
+ * constitution noise from near-perfect runs.
  */
 export async function accumulateReviewKnowledge(
   provider: AIProvider,
@@ -211,18 +249,29 @@ export async function accumulateReviewKnowledge(
 ): Promise<void> {
   console.log(chalk.blue("\n─── Knowledge Memory ────────────────────────────"));
 
+  // Extract review score for quality gate and lesson annotation
+  const reviewScoreMatch = reviewText.match(/Score:\s*(\d+(?:\.\d+)?)\s*\/\s*10/i);
+  const reviewScore = reviewScoreMatch ? parseFloat(reviewScoreMatch[1]) : undefined;
+
+  // Quality gate: excellent runs (≥ 9.0) rarely produce actionable lessons
+  if (reviewScore !== undefined && reviewScore >= 9.0) {
+    console.log(chalk.gray(`  Review score ${reviewScore}/10 — run quality excellent, skipping constitution update.`));
+    return;
+  }
+
   const issues = extractIssuesFromReview(reviewText);
   if (issues.length === 0) {
     console.log(chalk.gray("  No actionable issues found in review. Skipping."));
     return;
   }
 
-  console.log(chalk.gray(`  Extracted ${issues.length} issue(s) from review:`));
+  const scoreLabel = reviewScore !== undefined ? ` (review: ${reviewScore}/10)` : "";
+  console.log(chalk.gray(`  Extracted ${issues.length} issue(s) from review${scoreLabel}:`));
   for (const issue of issues) {
     console.log(chalk.gray(`    - [${issue.category}] ${issue.description.slice(0, 80)}`));
   }
 
-  await appendLessonsToConstitution(projectRoot, issues);
+  await appendLessonsToConstitution(projectRoot, issues, reviewScore);
 }
 
 // ─── Auto-Consolidation ──────────────────────────────────────────────────────