bmad-fh 6.0.0-alpha.23

package/tools/flattener/main.js

@@ -0,0 +1,483 @@
+const { Command } = require('commander');
+const fs = require('fs-extra');
+const path = require('node:path');
+const process = require('node:process');
+
+// Modularized components
+const { findProjectRoot } = require('./projectRoot.js');
+const { promptYesNo, promptPath } = require('./prompts.js');
+const { discoverFiles, filterFiles, aggregateFileContents } = require('./files.js');
+const { generateXMLOutput } = require('./xml.js');
+const { calculateStatistics } = require('./stats.js');
+
+/**
+ * Recursively discover all files in a directory
+ * @param {string} rootDir - The root directory to scan
+ * @returns {Promise<string[]>} Array of file paths
+ */
+
+/**
+ * Parse .gitignore file and return ignore patterns
+ * @param {string} gitignorePath - Path to .gitignore file
+ * @returns {Promise<string[]>} Array of ignore patterns
+ */
+
+/**
+ * Check if a file is binary using file command and heuristics
+ * @param {string} filePath - Path to the file
+ * @returns {Promise<boolean>} True if file is binary
+ */
+
+/**
+ * Read and aggregate content from text files
+ * @param {string[]} files - Array of file paths
+ * @param {string} rootDir - The root directory
+ * @param {Object} spinner - Optional spinner instance for progress display
+ * @returns {Promise<Object>} Object containing file contents and metadata
+ */
+
+/**
+ * Generate XML output with aggregated file contents using streaming
+ * @param {Object} aggregatedContent - The aggregated content object
+ * @param {string} outputPath - The output file path
+ * @returns {Promise<void>} Promise that resolves when writing is complete
+ */
+
+/**
+ * Calculate statistics for the processed files
+ * @param {Object} aggregatedContent - The aggregated content object
+ * @param {number} xmlFileSize - The size of the generated XML file in bytes
+ * @returns {Object} Statistics object
+ */
+
+/**
+ * Filter files based on .gitignore patterns
+ * @param {string[]} files - Array of file paths
+ * @param {string} rootDir - The root directory
+ * @returns {Promise<string[]>} Filtered array of file paths
+ */
+
+/**
+ * Attempt to find the project root by walking up from startDir
+ * Looks for common project markers like .git, package.json, pyproject.toml, etc.
+ * @param {string} startDir
+ * @returns {Promise<string|null>} project root directory or null if not found
+ */
+
+const program = new Command();
+
+program
+  .name('bmad-flatten')
+  .description('BMad-Method codebase flattener tool')
+  .version('1.0.0')
+  .option('-i, --input <path>', 'Input directory to flatten', process.cwd())
+  .option('-o, --output <path>', 'Output file path', 'flattened-codebase.xml')
+  .action(async (options) => {
+    let inputDir = path.resolve(options.input);
+    let outputPath = path.resolve(options.output);
+
+    // Detect if user explicitly provided -i/--input or -o/--output
+    const argv = process.argv.slice(2);
+    const userSpecifiedInput = argv.some((a) => a === '-i' || a === '--input' || a.startsWith('--input='));
+    const userSpecifiedOutput = argv.some((a) => a === '-o' || a === '--output' || a.startsWith('--output='));
+    const noPathArguments = !userSpecifiedInput && !userSpecifiedOutput;
+
+    if (noPathArguments) {
+      const detectedRoot = await findProjectRoot(process.cwd());
+      const suggestedOutput = detectedRoot ? path.join(detectedRoot, 'flattened-codebase.xml') : path.resolve('flattened-codebase.xml');
+
+      if (detectedRoot) {
+        const useDefaults = await promptYesNo(
+          `Detected project root at "${detectedRoot}". Use it as input and write output to "${suggestedOutput}"?`,
+          true,
+        );
+        if (useDefaults) {
+          inputDir = detectedRoot;
+          outputPath = suggestedOutput;
+        } else {
+          inputDir = await promptPath('Enter input directory path', process.cwd());
+          outputPath = await promptPath('Enter output file path', path.join(inputDir, 'flattened-codebase.xml'));
+        }
+      } else {
+        console.log('Could not auto-detect a project root.');
+        inputDir = await promptPath('Enter input directory path', process.cwd());
+        outputPath = await promptPath('Enter output file path', path.join(inputDir, 'flattened-codebase.xml'));
+      }
+    }
+
+    // Ensure output directory exists
+    await fs.ensureDir(path.dirname(outputPath));
+
+    try {
+      // Verify input directory exists
+      if (!(await fs.pathExists(inputDir))) {
+        console.error(`❌ Error: Input directory does not exist: ${inputDir}`);
+        process.exit(1);
+      }
+
+      // Import ora dynamically
+      const { default: ora } = await import('ora');
+
+      // Start file discovery with spinner
+      const discoverySpinner = ora('🔍 Discovering files...').start();
+      const files = await discoverFiles(inputDir);
+      const filteredFiles = await filterFiles(files, inputDir);
+      discoverySpinner.succeed(`📁 Found ${filteredFiles.length} files to include`);
+
+      // Process files with progress tracking
+      console.log('Reading file contents');
+      const processingSpinner = ora('📄 Processing files...').start();
+      const aggregatedContent = await aggregateFileContents(filteredFiles, inputDir, processingSpinner);
+      processingSpinner.succeed(`✅ Processed ${aggregatedContent.processedFiles}/${filteredFiles.length} files`);
+      if (aggregatedContent.errors.length > 0) {
+        console.log(`Errors: ${aggregatedContent.errors.length}`);
+      }
+
+      // Generate XML output using streaming
+      const xmlSpinner = ora('🔧 Generating XML output...').start();
+      await generateXMLOutput(aggregatedContent, outputPath);
+      xmlSpinner.succeed('📝 XML generation completed');
+
+      // Calculate and display statistics
+      const outputStats = await fs.stat(outputPath);
+      const stats = await calculateStatistics(aggregatedContent, outputStats.size, inputDir);
+
+      // Display completion summary
+      console.log('\n📊 Completion Summary:');
+      console.log(`✅ Successfully processed ${filteredFiles.length} files into ${path.basename(outputPath)}`);
+      console.log(`📁 Output file: ${outputPath}`);
+      console.log(`📏 Total source size: ${stats.totalSize}`);
+      console.log(`📄 Generated XML size: ${stats.xmlSize}`);
+      console.log(`📝 Total lines of code: ${stats.totalLines.toLocaleString()}`);
+      console.log(`🔢 Estimated tokens: ${stats.estimatedTokens}`);
+      console.log(`📊 File breakdown: ${stats.textFiles} text, ${stats.binaryFiles} binary, ${stats.errorFiles} errors\n`);
+
+      // Ask user if they want detailed stats + markdown report
+      const generateDetailed = await promptYesNo('Generate detailed stats (console + markdown) now?', true);
+
+      if (generateDetailed) {
+        // Additional detailed stats
+        console.log('\n📈 Size Percentiles:');
+        console.log(
+          `   Avg: ${Math.round(stats.avgFileSize).toLocaleString()} B, Median: ${Math.round(
+            stats.medianFileSize,
+          ).toLocaleString()} B, p90: ${stats.p90.toLocaleString()} B, p95: ${stats.p95.toLocaleString()} B, p99: ${stats.p99.toLocaleString()} B`,
+        );
+
+        if (Array.isArray(stats.histogram) && stats.histogram.length > 0) {
+          console.log('\n🧮 Size Histogram:');
+          for (const b of stats.histogram.slice(0, 2)) {
+            console.log(`   ${b.label}: ${b.count} files, ${b.bytes.toLocaleString()} bytes`);
+          }
+          if (stats.histogram.length > 2) {
+            console.log(`   … and ${stats.histogram.length - 2} more buckets`);
+          }
+        }
+
+        if (Array.isArray(stats.byExtension) && stats.byExtension.length > 0) {
+          const topExt = stats.byExtension.slice(0, 2);
+          console.log('\n📦 Top Extensions:');
+          for (const e of topExt) {
+            const pct = stats.totalBytes ? (e.bytes / stats.totalBytes) * 100 : 0;
+            console.log(`   ${e.ext}: ${e.count} files, ${e.bytes.toLocaleString()} bytes (${pct.toFixed(2)}%)`);
+          }
+          if (stats.byExtension.length > 2) {
+            console.log(`   … and ${stats.byExtension.length - 2} more extensions`);
+          }
+        }
+
+        if (Array.isArray(stats.byDirectory) && stats.byDirectory.length > 0) {
+          const topDir = stats.byDirectory.slice(0, 2);
+          console.log('\n📂 Top Directories:');
+          for (const d of topDir) {
+            const pct = stats.totalBytes ? (d.bytes / stats.totalBytes) * 100 : 0;
+            console.log(`   ${d.dir}: ${d.count} files, ${d.bytes.toLocaleString()} bytes (${pct.toFixed(2)}%)`);
+          }
+          if (stats.byDirectory.length > 2) {
+            console.log(`   … and ${stats.byDirectory.length - 2} more directories`);
+          }
+        }
+
+        if (Array.isArray(stats.depthDistribution) && stats.depthDistribution.length > 0) {
+          console.log('\n🌳 Depth Distribution:');
+          const dd = stats.depthDistribution.slice(0, 2);
+          let line = '   ' + dd.map((d) => `${d.depth}:${d.count}`).join('  ');
+          if (stats.depthDistribution.length > 2) {
+            line += `  … +${stats.depthDistribution.length - 2} more`;
+          }
+          console.log(line);
+        }
+
+        if (Array.isArray(stats.longestPaths) && stats.longestPaths.length > 0) {
+          console.log('\n🧵 Longest Paths:');
+          for (const p of stats.longestPaths.slice(0, 2)) {
+            console.log(`   ${p.path} (${p.length} chars, ${p.size.toLocaleString()} bytes)`);
+          }
+          if (stats.longestPaths.length > 2) {
+            console.log(`   … and ${stats.longestPaths.length - 2} more paths`);
+          }
+        }
+
+        if (stats.temporal) {
+          console.log('\n⏱️ Temporal:');
+          if (stats.temporal.oldest) {
+            console.log(`   Oldest: ${stats.temporal.oldest.path} (${stats.temporal.oldest.mtime})`);
+          }
+          if (stats.temporal.newest) {
+            console.log(`   Newest: ${stats.temporal.newest.path} (${stats.temporal.newest.mtime})`);
+          }
+          if (Array.isArray(stats.temporal.ageBuckets)) {
+            console.log('   Age buckets:');
+            for (const b of stats.temporal.ageBuckets.slice(0, 2)) {
+              console.log(`     ${b.label}: ${b.count} files, ${b.bytes.toLocaleString()} bytes`);
+            }
+            if (stats.temporal.ageBuckets.length > 2) {
+              console.log(`     … and ${stats.temporal.ageBuckets.length - 2} more buckets`);
+            }
+          }
+        }
+
+        if (stats.quality) {
+          console.log('\n✅ Quality Signals:');
+          console.log(`   Zero-byte files: ${stats.quality.zeroByteFiles}`);
+          console.log(`   Empty text files: ${stats.quality.emptyTextFiles}`);
+          console.log(`   Hidden files: ${stats.quality.hiddenFiles}`);
+          console.log(`   Symlinks: ${stats.quality.symlinks}`);
+          console.log(
+            `   Large files (>= ${(stats.quality.largeThreshold / (1024 * 1024)).toFixed(0)} MB): ${stats.quality.largeFilesCount}`,
+          );
+          console.log(`   Suspiciously large files (>= 100 MB): ${stats.quality.suspiciousLargeFilesCount}`);
+        }
+
+        if (Array.isArray(stats.duplicateCandidates) && stats.duplicateCandidates.length > 0) {
+          console.log('\n🧬 Duplicate Candidates:');
+          for (const d of stats.duplicateCandidates.slice(0, 2)) {
+            console.log(`   ${d.reason}: ${d.count} files @ ${d.size.toLocaleString()} bytes`);
+          }
+          if (stats.duplicateCandidates.length > 2) {
+            console.log(`   … and ${stats.duplicateCandidates.length - 2} more groups`);
+          }
+        }
+
+        if (typeof stats.compressibilityRatio === 'number') {
+          console.log(`\n🗜️ Compressibility ratio (sampled): ${(stats.compressibilityRatio * 100).toFixed(2)}%`);
+        }
+
+        if (stats.git && stats.git.isRepo) {
+          console.log('\n🔧 Git:');
+          console.log(`   Tracked: ${stats.git.trackedCount} files, ${stats.git.trackedBytes.toLocaleString()} bytes`);
+          console.log(`   Untracked: ${stats.git.untrackedCount} files, ${stats.git.untrackedBytes.toLocaleString()} bytes`);
+          if (Array.isArray(stats.git.lfsCandidates) && stats.git.lfsCandidates.length > 0) {
+            console.log('   LFS candidates (top 2):');
+            for (const f of stats.git.lfsCandidates.slice(0, 2)) {
+              console.log(`     ${f.path} (${f.size.toLocaleString()} bytes)`);
+            }
+            if (stats.git.lfsCandidates.length > 2) {
+              console.log(`     … and ${stats.git.lfsCandidates.length - 2} more`);
+            }
+          }
+        }
+
+        if (Array.isArray(stats.largestFiles) && stats.largestFiles.length > 0) {
+          console.log('\n📚 Largest Files (top 2):');
+          for (const f of stats.largestFiles.slice(0, 2)) {
+            // Show LOC for text files when available; omit ext and mtime
+            let locStr = '';
+            if (!f.isBinary && Array.isArray(aggregatedContent?.textFiles)) {
+              const tf = aggregatedContent.textFiles.find((t) => t.path === f.path);
+              if (tf && typeof tf.lines === 'number') {
+                locStr = `, LOC: ${tf.lines.toLocaleString()}`;
+              }
+            }
+            console.log(`   ${f.path} – ${f.sizeFormatted} (${f.percentOfTotal.toFixed(2)}%)${locStr}`);
+          }
+          if (stats.largestFiles.length > 2) {
+            console.log(`   … and ${stats.largestFiles.length - 2} more files`);
+          }
+        }
+
+        // Write a comprehensive markdown report next to the XML
+        {
+          const mdPath = outputPath.endsWith('.xml') ? outputPath.replace(/\.xml$/i, '.stats.md') : outputPath + '.stats.md';
+          try {
+            const pct = (num, den) => (den ? (num / den) * 100 : 0);
+            const md = [];
+            md.push(
+              `# 🧾 Flatten Stats for ${path.basename(outputPath)}`,
+              '',
+              '## 📊 Summary',
+              `- Total source size: ${stats.totalSize}`,
+              `- Generated XML size: ${stats.xmlSize}`,
+              `- Total lines of code: ${stats.totalLines.toLocaleString()}`,
+              `- Estimated tokens: ${stats.estimatedTokens}`,
+              `- File breakdown: ${stats.textFiles} text, ${stats.binaryFiles} binary, ${stats.errorFiles} errors`,
+              '',
+              '## 📈 Size Percentiles',
+              `Avg: ${Math.round(stats.avgFileSize).toLocaleString()} B, Median: ${Math.round(
+                stats.medianFileSize,
+              ).toLocaleString()} B, p90: ${stats.p90.toLocaleString()} B, p95: ${stats.p95.toLocaleString()} B, p99: ${stats.p99.toLocaleString()} B`,
+              '',
+            );
+
+            // Histogram
+            if (Array.isArray(stats.histogram) && stats.histogram.length > 0) {
+              md.push('## 🧮 Size Histogram', '| Bucket | Files | Bytes |', '| --- | ---: | ---: |');
+              for (const b of stats.histogram) {
+                md.push(`| ${b.label} | ${b.count} | ${b.bytes.toLocaleString()} |`);
+              }
+              md.push('');
+            }
+
+            // Top Extensions
+            if (Array.isArray(stats.byExtension) && stats.byExtension.length > 0) {
+              md.push('## 📦 Top Extensions by Bytes (Top 20)', '| Ext | Files | Bytes | % of total |', '| --- | ---: | ---: | ---: |');
+              for (const e of stats.byExtension.slice(0, 20)) {
+                const p = pct(e.bytes, stats.totalBytes);
+                md.push(`| ${e.ext} | ${e.count} | ${e.bytes.toLocaleString()} | ${p.toFixed(2)}% |`);
+              }
+              md.push('');
+            }
+
+            // Top Directories
+            if (Array.isArray(stats.byDirectory) && stats.byDirectory.length > 0) {
+              md.push(
+                '## 📂 Top Directories by Bytes (Top 20)',
+                '| Directory | Files | Bytes | % of total |',
+                '| --- | ---: | ---: | ---: |',
+              );
+              for (const d of stats.byDirectory.slice(0, 20)) {
+                const p = pct(d.bytes, stats.totalBytes);
+                md.push(`| ${d.dir} | ${d.count} | ${d.bytes.toLocaleString()} | ${p.toFixed(2)}% |`);
+              }
+              md.push('');
+            }
+
+            // Depth distribution
+            if (Array.isArray(stats.depthDistribution) && stats.depthDistribution.length > 0) {
+              md.push('## 🌳 Depth Distribution', '| Depth | Count |', '| ---: | ---: |');
+              for (const d of stats.depthDistribution) {
+                md.push(`| ${d.depth} | ${d.count} |`);
+              }
+              md.push('');
+            }
+
+            // Longest paths
+            if (Array.isArray(stats.longestPaths) && stats.longestPaths.length > 0) {
+              md.push('## 🧵 Longest Paths (Top 25)', '| Path | Length | Bytes |', '| --- | ---: | ---: |');
+              for (const pth of stats.longestPaths) {
+                md.push(`| ${pth.path} | ${pth.length} | ${pth.size.toLocaleString()} |`);
+              }
+              md.push('');
+            }
+
+            // Temporal
+            if (stats.temporal) {
+              md.push('## ⏱️ Temporal');
+              if (stats.temporal.oldest) {
+                md.push(`- Oldest: ${stats.temporal.oldest.path} (${stats.temporal.oldest.mtime})`);
+              }
+              if (stats.temporal.newest) {
+                md.push(`- Newest: ${stats.temporal.newest.path} (${stats.temporal.newest.mtime})`);
+              }
+              if (Array.isArray(stats.temporal.ageBuckets)) {
+                md.push('', '| Age | Files | Bytes |', '| --- | ---: | ---: |');
+                for (const b of stats.temporal.ageBuckets) {
+                  md.push(`| ${b.label} | ${b.count} | ${b.bytes.toLocaleString()} |`);
+                }
+              }
+              md.push('');
+            }
+
+            // Quality signals
+            if (stats.quality) {
+              md.push(
+                '## ✅ Quality Signals',
+                `- Zero-byte files: ${stats.quality.zeroByteFiles}`,
+                `- Empty text files: ${stats.quality.emptyTextFiles}`,
+                `- Hidden files: ${stats.quality.hiddenFiles}`,
+                `- Symlinks: ${stats.quality.symlinks}`,
+                `- Large files (>= ${(stats.quality.largeThreshold / (1024 * 1024)).toFixed(0)} MB): ${stats.quality.largeFilesCount}`,
+                `- Suspiciously large files (>= 100 MB): ${stats.quality.suspiciousLargeFilesCount}`,
+                '',
+              );
+            }
+
+            // Duplicates
+            if (Array.isArray(stats.duplicateCandidates) && stats.duplicateCandidates.length > 0) {
+              md.push('## 🧬 Duplicate Candidates', '| Reason | Files | Size (bytes) |', '| --- | ---: | ---: |');
+              for (const d of stats.duplicateCandidates) {
+                md.push(`| ${d.reason} | ${d.count} | ${d.size.toLocaleString()} |`);
+              }
+              md.push('', '### 🧬 Duplicate Groups Details');
+              let dupIndex = 1;
+              for (const d of stats.duplicateCandidates) {
+                md.push(`#### Group ${dupIndex}: ${d.count} files @ ${d.size.toLocaleString()} bytes (${d.reason})`);
+                if (Array.isArray(d.files) && d.files.length > 0) {
+                  for (const fp of d.files) {
+                    md.push(`- ${fp}`);
+                  }
+                } else {
+                  md.push('- (file list unavailable)');
+                }
+                md.push('');
+                dupIndex++;
+              }
+              md.push('');
+            }
+
+            // Compressibility
+            if (typeof stats.compressibilityRatio === 'number') {
+              md.push('## 🗜️ Compressibility', `Sampled compressibility ratio: ${(stats.compressibilityRatio * 100).toFixed(2)}%`, '');
+            }
+
+            // Git
+            if (stats.git && stats.git.isRepo) {
+              md.push(
+                '## 🔧 Git',
+                `- Tracked: ${stats.git.trackedCount} files, ${stats.git.trackedBytes.toLocaleString()} bytes`,
+                `- Untracked: ${stats.git.untrackedCount} files, ${stats.git.untrackedBytes.toLocaleString()} bytes`,
+              );
+              if (Array.isArray(stats.git.lfsCandidates) && stats.git.lfsCandidates.length > 0) {
+                md.push('', '### 📦 LFS Candidates (Top 20)', '| Path | Bytes |', '| --- | ---: |');
+                for (const f of stats.git.lfsCandidates.slice(0, 20)) {
+                  md.push(`| ${f.path} | ${f.size.toLocaleString()} |`);
+                }
+              }
+              md.push('');
+            }
+
+            // Largest Files
+            if (Array.isArray(stats.largestFiles) && stats.largestFiles.length > 0) {
+              md.push('## 📚 Largest Files (Top 50)', '| Path | Size | % of total | LOC |', '| --- | ---: | ---: | ---: |');
+              for (const f of stats.largestFiles) {
+                let loc = '';
+                if (!f.isBinary && Array.isArray(aggregatedContent?.textFiles)) {
+                  const tf = aggregatedContent.textFiles.find((t) => t.path === f.path);
+                  if (tf && typeof tf.lines === 'number') {
+                    loc = tf.lines.toLocaleString();
+                  }
+                }
+                md.push(`| ${f.path} | ${f.sizeFormatted} | ${f.percentOfTotal.toFixed(2)}% | ${loc} |`);
+              }
+              md.push('');
+            }
+
+            await fs.writeFile(mdPath, md.join('\n'));
+            console.log(`\n🧾 Detailed stats report written to: ${mdPath}`);
+          } catch (error) {
+            console.warn(`⚠️ Failed to write stats markdown: ${error.message}`);
+          }
+        }
+      }
+    } catch (error) {
+      console.error('❌ Critical error:', error.message);
+      console.error('An unexpected error occurred.');
+      process.exit(1);
+    }
+  });
+
+if (require.main === module) {
+  program.parse();
+}
+
+module.exports = program;

package/tools/flattener/projectRoot.js

@@ -0,0 +1,201 @@
+const fs = require('fs-extra');
+const path = require('node:path');
+
+// Deno/Node compatibility: explicitly import process
+const process = require('node:process');
+const { execFile } = require('node:child_process');
+const { promisify } = require('node:util');
+const execFileAsync = promisify(execFile);
+
+// Simple memoization across calls (keyed by realpath of startDir)
+const _cache = new Map();
+
+async function _tryRun(cmd, args, cwd, timeoutMs = 500) {
+  try {
+    const { stdout } = await execFileAsync(cmd, args, {
+      cwd,
+      timeout: timeoutMs,
+      windowsHide: true,
+      maxBuffer: 1024 * 1024,
+    });
+    const out = String(stdout || '').trim();
+    return out || null;
+  } catch {
+    return null;
+  }
+}
+
+async function _detectVcsTopLevel(startDir) {
+  // Run common VCS root queries in parallel; ignore failures
+  const gitP = _tryRun('git', ['rev-parse', '--show-toplevel'], startDir);
+  const hgP = _tryRun('hg', ['root'], startDir);
+  const svnP = (async () => {
+    const show = await _tryRun('svn', ['info', '--show-item', 'wc-root'], startDir);
+    if (show) return show;
+    const info = await _tryRun('svn', ['info'], startDir);
+    if (info) {
+      const line = info.split(/\r?\n/).find((l) => l.toLowerCase().startsWith('working copy root path:'));
+      if (line) return line.split(':').slice(1).join(':').trim();
+    }
+    return null;
+  })();
+  const [git, hg, svn] = await Promise.all([gitP, hgP, svnP]);
+  return git || hg || svn || null;
+}
+
+/**
+ * Attempt to find the project root by walking up from startDir.
+ * Uses a robust, prioritized set of ecosystem markers (VCS > workspaces/monorepo > lock/build > language config).
+ * Also recognizes package.json with "workspaces" as a workspace root.
+ * You can augment markers via env PROJECT_ROOT_MARKERS as a comma-separated list of file/dir names.
+ * @param {string} startDir
+ * @returns {Promise<string|null>} project root directory or null if not found
+ */
+async function findProjectRoot(startDir) {
+  try {
+    // Resolve symlinks for robustness (e.g., when invoked from a symlinked path)
+    let dir = path.resolve(startDir);
+    try {
+      dir = await fs.realpath(dir);
+    } catch {
+      // ignore if realpath fails; continue with resolved path
+    }
+    const startKey = dir; // preserve starting point for caching
+    if (_cache.has(startKey)) return _cache.get(startKey);
+    const fsRoot = path.parse(dir).root;
+
+    // Helper to safely check for existence
+    const exists = (p) => fs.pathExists(p);
+
+    // Build checks: an array of { makePath: (dir) => string, weight }
+    const checks = [];
+
+    const add = (rel, weight) => {
+      const makePath = (d) => (Array.isArray(rel) ? path.join(d, ...rel) : path.join(d, rel));
+      checks.push({ makePath, weight });
+    };
+
+    // Highest priority: explicit sentinel markers
+    add('.project-root', 110);
+    add('.workspace-root', 110);
+    add('.repo-root', 110);
+
+    // Highest priority: VCS roots
+    add('.git', 100);
+    add('.hg', 95);
+    add('.svn', 95);
+
+    // Monorepo/workspace indicators
+    add('pnpm-workspace.yaml', 90);
+    add('lerna.json', 90);
+    add('turbo.json', 90);
+    add('nx.json', 90);
+    add('rush.json', 90);
+    add('go.work', 90);
+    add('WORKSPACE', 90);
+    add('WORKSPACE.bazel', 90);
+    add('MODULE.bazel', 90);
+    add('pants.toml', 90);
+
+    // Lockfiles and package-manager/top-level locks
+    add('yarn.lock', 85);
+    add('pnpm-lock.yaml', 85);
+    add('package-lock.json', 85);
+    add('bun.lockb', 85);
+    add('Cargo.lock', 85);
+    add('composer.lock', 85);
+    add('poetry.lock', 85);
+    add('Pipfile.lock', 85);
+    add('Gemfile.lock', 85);
+
+    // Build-system root indicators
+    add('settings.gradle', 80);
+    add('settings.gradle.kts', 80);
+    add('gradlew', 80);
+    add('pom.xml', 80);
+    add('build.sbt', 80);
+    add(['project', 'build.properties'], 80);
+
+    // Language/project config markers
+    add('deno.json', 75);
+    add('deno.jsonc', 75);
+    add('pyproject.toml', 75);
+    add('Pipfile', 75);
+    add('requirements.txt', 75);
+    add('go.mod', 75);
+    add('Cargo.toml', 75);
+    add('composer.json', 75);
+    add('mix.exs', 75);
+    add('Gemfile', 75);
+    add('CMakeLists.txt', 75);
+    add('stack.yaml', 75);
+    add('cabal.project', 75);
+    add('rebar.config', 75);
+    add('pubspec.yaml', 75);
+    add('flake.nix', 75);
+    add('shell.nix', 75);
+    add('default.nix', 75);
+    add('.tool-versions', 75);
+    add('package.json', 74); // generic Node project (lower than lockfiles/workspaces)
+
+    // Changesets
+    add(['.changeset', 'config.json'], 70);
+    add('.changeset', 70);
+
+    // Custom markers via env (comma-separated names)
+    if (process.env.PROJECT_ROOT_MARKERS) {
+      for (const name of process.env.PROJECT_ROOT_MARKERS.split(',')
+        .map((s) => s.trim())
+        .filter(Boolean)) {
+        add(name, 72);
+      }
+    }
+
+    /** Check for package.json with "workspaces" */
+    const hasWorkspacePackageJson = async (d) => {
+      const pkgPath = path.join(d, 'package.json');
+      if (!(await exists(pkgPath))) return false;
+      try {
+        const raw = await fs.readFile(pkgPath, 'utf8');
+        const pkg = JSON.parse(raw);
+        return Boolean(pkg && pkg.workspaces);
+      } catch {
+        return false;
+      }
+    };
+
+    let best = null; // { dir, weight }
+
+    // Try to detect VCS toplevel once up-front; treat as authoritative slightly above .git marker
+    const vcsTop = await _detectVcsTopLevel(dir);
+    if (vcsTop) {
+      best = { dir: vcsTop, weight: 101 };
+    }
+
+    while (true) {
+      // Special check: package.json with "workspaces"
+      if ((await hasWorkspacePackageJson(dir)) && (!best || 90 >= best.weight)) best = { dir, weight: 90 };
+
+      // Evaluate all other checks in parallel
+      const results = await Promise.all(checks.map(async (c) => ({ c, ok: await exists(c.makePath(dir)) })));
+
+      for (const { c, ok } of results) {
+        if (!ok) continue;
+        if (!best || c.weight >= best.weight) {
+          best = { dir, weight: c.weight };
+        }
+      }
+
+      if (dir === fsRoot) break;
+      dir = path.dirname(dir);
+    }
+
+    const out = best ? best.dir : null;
+    _cache.set(startKey, out);
+    return out;
+  } catch {
+    return null;
+  }
+}
+
+module.exports = { findProjectRoot };