ctxman 3.0.0 → 3.1.0

package/lib/analyzers/token-calculator.js

@@ -1,577 +1,968 @@
 import fs from 'fs';
 import path from 'path';
-import { execSync } from 'child_process';
 import readline from 'readline';
 import TokenUtils from '../utils/token-utils.js';
 import FileUtils from '../utils/file-utils.js';
 import ClipboardUtils from '../utils/clipboard-utils.js';
 import ConfigUtils from '../utils/config-utils.js';
-import GitIgnoreParser from '../parsers/gitignore-parser.js';
 import MethodAnalyzer from './method-analyzer.js';
-import MethodFilterParser from '../parsers/method-filter-parser.js';
 import GitIngestFormatter from '../formatters/gitingest-formatter.js';
+import FormatRegistry from '../formatters/format-registry.js';
+import { formatRedactionSummary, redactSecrets } from '../utils/secret-redactor.js';
 import { LLMDetector } from '../utils/llm-detector.js';
+import { POLICY_FILE } from '../policy/policy.js';
+import { enforceExport, PolicyViolationError } from '../policy/enforce.js';
+import { recordExport } from '../policy/audit.js';
+import { loadIgnoreFileRules, resolveIgnoreFileNames } from '../policy/ignore-files.js';
+
+// What ctxman itself writes into a project: reports, context exports and digests in the root,
+// and its state under .ctxman/. A later run must not count them as project files.
+const OWN_OUTPUT_PATHS = [
+  /^token-analysis-report\.json$/,
+  /^llm-context\.\w+$/,
+  /^digest(-chunk-\d+-of-\d+)?\.txt$/,
+  /^(ai-suggestions-report|dependency-context|restored-context)\.json$/,
+  /^performance-report\.txt$/,
+  /^\.ctxman[/\\]audit\.log$/,
+  /^\.ctxman[/\\](versions|snapshots|metrics|logs|temp|self-knowledge|rag-store)$/,
+];
 
 class TokenCalculator {
-    constructor(projectRoot, options = {}) {
-        this.projectRoot = projectRoot;
-        this.options = { verbose: false, compactContext: true, methodLevel: false, ...options };
-        this.stats = this.initStats();
-        this.gitIgnore = this.initGitIgnore();
-        this.methodAnalyzer = new MethodAnalyzer();
-        this.methodFilter = this.options.methodLevel ? this.initMethodFilter() : null;
-        this.methodStats = { totalMethods: 0, includedMethods: 0, methodTokens: {} };
+  constructor(projectRoot, options = {}) {
+    this.projectRoot = projectRoot;
+    this.options = { verbose: false, compactContext: true, methodLevel: false, ...options };
+    this.stats = this.initStats();
+    this.gitIgnore = this.initGitIgnore();
+    this.patternFilters = this.initPatternFilters();
+    this.methodAnalyzer = new MethodAnalyzer();
+    this.methodFilter = this.options.methodLevel ? this.initMethodFilter() : null;
+    this.methodStats = { totalMethods: 0, includedMethods: 0, methodTokens: {} };
+    // o200k_base for OpenAI models since GPT-4o, cl100k_base otherwise
+    this.encoding = TokenUtils.encodingFor(this.options.targetModel);
+
+    // Export policy from loadPolicy() (lib/policy): deny rules, export limits, audit log.
+    // Without one, scanning and exporting behave as before.
+    this.policy = this.options.policy?.active ? this.options.policy : null;
+    this.ignoreFileRules = this.initIgnoreFileRules();
+    if (this.policy) this.stats.policyDeniedFiles = 0;
+    if (this.ignoreFileRules) this.stats.ignoreFileIgnoredFiles = 0;
+  }
+
+  /**
+   * AI-tool ignore files (.cursorignore, .aiexclude, ...) from --respect-ignore-files and
+   * the policy's ignoreFiles
+   * @returns {import('../policy/ignore-files.js').IgnoreFileRules|null}
+   */
+  initIgnoreFileRules() {
+    const names = [
+      ...new Set([
+        ...resolveIgnoreFileNames(this.options.respectIgnoreFiles),
+        ...(this.policy?.ignoreFiles ?? []),
+      ]),
+    ];
+    return names.length > 0 ? loadIgnoreFileRules(this.projectRoot, names) : null;
+  }
+
+  initMethodFilter() {
+    return ConfigUtils.initMethodFilter(this.projectRoot);
+  }
+
+  initStats() {
+    return {
+      totalFiles: 0,
+      totalTokens: 0,
+      totalBytes: 0,
+      totalLines: 0,
+      ignoredFiles: 0,
+      calculatorIgnoredFiles: 0,
+      byExtension: {},
+      byDirectory: {},
+      largestFiles: [],
+    };
+  }
+
+  initGitIgnore() {
+    return ConfigUtils.initGitIgnore(this.projectRoot);
+  }
+
+  /**
+   * Include/exclude patterns of --template and --profile, with .contextinclude /
+   * .contextignore syntax. A file must pass every active set.
+   */
+  initPatternFilters() {
+    const toRegex = (pattern) => this.gitIgnore.convertToRegex(pattern);
+    return [this.options.templateConfig, this.options.profileConfig]
+      .filter(Boolean)
+      .map((config) => ({
+        include: (config.includePatterns || []).map(toRegex),
+        exclude: (config.ignorePatterns || []).map(toRegex),
+      }));
+  }
+
+  passesPatternFilters(relativePath) {
+    return this.patternFilters.every(({ include, exclude }) => {
+      if (include.length > 0 && !this.gitIgnore.testPatternsWithNegation(include, relativePath)) {
+        return false;
+      }
+      // Last matching pattern wins, as in .contextignore
+      let excluded = false;
+      for (const pattern of exclude) {
+        if (pattern.regex.test(relativePath)) excluded = !pattern.isNegation;
+      }
+      return !excluded;
+    });
+  }
+
+  calculateTokens(content, filePath) {
+    return TokenUtils.calculate(content, filePath, this.encoding);
+  }
+
+  isTextFile(filePath) {
+    return FileUtils.isText(filePath);
+  }
+
+  analyzeFile(filePath) {
+    try {
+      const content = fs.readFileSync(filePath, 'utf8');
+      const stats = fs.statSync(filePath);
+
+      const fileInfo = {
+        path: filePath,
+        relativePath: path.relative(this.projectRoot, filePath),
+        sizeBytes: stats.size,
+        tokens: this.calculateTokens(content, filePath),
+        lines: content.split('\n').length,
+        extension: path.extname(filePath).toLowerCase() || 'no-extension',
+      };
+
+      // Method-level analysis if enabled
+      if (this.options.methodLevel && this.isCodeFile(filePath)) {
+        fileInfo.methods = this.analyzeFileMethods(content, filePath);
+      }
+
+      return fileInfo;
+    } catch (error) {
+      return {
+        path: filePath,
+        relativePath: path.relative(this.projectRoot, filePath),
+        sizeBytes: 0,
+        tokens: 0,
+        lines: 0,
+        extension: 'error',
+        error: error.message,
+      };
     }
-
-    initMethodFilter() {
-        return ConfigUtils.initMethodFilter(this.projectRoot);
-    }
-
-    initStats() {
-        return {
-            totalFiles: 0, totalTokens: 0, totalBytes: 0, totalLines: 0,
-            ignoredFiles: 0, calculatorIgnoredFiles: 0,
-            byExtension: {}, byDirectory: {}, largestFiles: []
+  }
+
+  isCodeFile(filePath) {
+    return FileUtils.isCode(filePath);
+  }
+
+  analyzeFileMethods(content, filePath) {
+    const methods = this.methodAnalyzer.extractMethods(content, filePath);
+    const sources = this.methodAnalyzer.extractMethodSources(content, methods, filePath);
+    const fileName = path.basename(filePath, path.extname(filePath));
+    const filteredMethods = [];
+
+    this.methodStats.totalMethods += methods.length;
+
+    methods.forEach((method, i) => {
+      if (!this.methodFilter || this.methodFilter.shouldIncludeMethod(method.name, fileName)) {
+        // `content` keeps the narrower legacy extraction that reports have always carried
+        const methodContent = this.methodAnalyzer.extractMethodContent(content, method.name);
+        const methodTokens = this.calculateTokens(sources[i], filePath);
+
+        const methodInfo = {
+          ...method,
+          tokens: methodTokens,
+          content: methodContent,
         };
-    }
-
-    initGitIgnore() {
-        return ConfigUtils.initGitIgnore(this.projectRoot);
-    }
-
-    calculateTokens(content, filePath) {
-        const tokens = TokenUtils.calculate(content, filePath);
-        return tokens;
-    }
-
-    isTextFile(filePath) {
-        return FileUtils.isText(filePath);
-    }
 
-    analyzeFile(filePath) {
-        try {
-            const content = fs.readFileSync(filePath, 'utf8');
-            const stats = fs.statSync(filePath);
-
-            const fileInfo = {
-                path: filePath,
-                relativePath: path.relative(this.projectRoot, filePath),
-                sizeBytes: stats.size,
-                tokens: this.calculateTokens(content, filePath),
-                lines: content.split('\n').length,
-                extension: path.extname(filePath).toLowerCase() || 'no-extension'
-            };
-
-            // Method-level analysis if enabled
-            if (this.options.methodLevel && this.isCodeFile(filePath)) {
-                fileInfo.methods = this.analyzeFileMethods(content, filePath);
-            }
-
-            return fileInfo;
-        } catch (error) {
-            return {
-                path: filePath,
-                relativePath: path.relative(this.projectRoot, filePath),
-                sizeBytes: 0, tokens: 0, lines: 0,
-                extension: 'error', error: error.message
-            };
-        }
-    }
-
-    isCodeFile(filePath) {
-        return FileUtils.isCode(filePath);
-    }
+        filteredMethods.push(methodInfo);
+        this.methodStats.includedMethods++;
+        this.methodStats.methodTokens[`${fileName}.${method.name}`] = methodTokens;
+      }
+    });
 
-    analyzeFileMethods(content, filePath) {
-        const methods = this.methodAnalyzer.extractMethods(content, filePath);
-        const fileName = path.basename(filePath, path.extname(filePath));
-        const filteredMethods = [];
+    return filteredMethods;
+  }
 
-        this.methodStats.totalMethods += methods.length;
+  scanDirectory(dir) {
+    const files = [];
 
-        for (const method of methods) {
-            if (!this.methodFilter || this.methodFilter.shouldIncludeMethod(method.name, fileName)) {
-                const methodContent = this.methodAnalyzer.extractMethodContent(content, method.name);
-                const methodTokens = methodContent ? this.calculateTokens(methodContent, filePath) : 0;
+    try {
+      for (const item of fs.readdirSync(dir)) {
+        const fullPath = path.join(dir, item);
+        const relativePath = path.relative(this.projectRoot, fullPath);
 
-                const methodInfo = {
-                    ...method,
-                    tokens: methodTokens,
-                    content: methodContent
-                };
-
-                filteredMethods.push(methodInfo);
-                this.methodStats.includedMethods++;
-                this.methodStats.methodTokens[`${fileName}.${method.name}`] = methodTokens;
-            }
+        if (this.gitIgnore.isIgnored(fullPath, relativePath)) {
+          this.countIgnoredFiles(fullPath);
+          continue;
         }
 
-        return filteredMethods;
-    }
-
-    scanDirectory(dir) {
-        const files = [];
-
-        try {
-            for (const item of fs.readdirSync(dir)) {
-                const fullPath = path.join(dir, item);
-                const relativePath = path.relative(this.projectRoot, fullPath);
-
-                if (this.gitIgnore.isIgnored(fullPath, relativePath)) {
-                    this.countIgnoredFiles(fullPath);
-                    continue;
-                }
-
-                const stat = fs.statSync(fullPath);
-                if (stat.isDirectory()) {
-                    if (!['node_modules', '.git', '.svn', '.hg', 'coverage', 'dist', 'build'].includes(item)) {
-                        files.push(...this.scanDirectory(fullPath));
-                    }
-                } else if (this.isTextFile(fullPath)) {
-                    files.push(fullPath);
-                }
-            }
-        } catch (error) {
-            console.error(`Error scanning directory ${dir}:`, error.message);
+        if (this.isOwnOutput(relativePath)) {
+          continue;
         }
 
-        return files;
-    }
-
-    countIgnoredFiles(filePath) {
-        const isCalculatorIgnored = ['calculator', 'calculator-include'].includes(
-            this.gitIgnore._lastIgnoreReason
-        );
-
-        try {
-            const stat = fs.statSync(filePath);
-            if (stat.isDirectory()) {
-                const count = this.countFilesInDirectory(filePath);
-                if (isCalculatorIgnored) {
-                    this.stats.calculatorIgnoredFiles += count;
-                } else {
-                    this.stats.ignoredFiles += count;
-                }
-            } else {
-                if (isCalculatorIgnored) {
-                    this.stats.calculatorIgnoredFiles++;
-                } else {
-                    this.stats.ignoredFiles++;
-                }
-            }
-        } catch { }
-    }
+        if (this.ignoreFileRules?.isIgnored(relativePath)) {
+          this.stats.ignoreFileIgnoredFiles += this.countMatchedFiles(fullPath);
+          continue;
+        }
 
-    countFilesInDirectory(dirPath) {
-        let count = 0;
-        try {
-            for (const item of fs.readdirSync(dirPath)) {
-                const fullPath = path.join(dirPath, item);
-                const stat = fs.statSync(fullPath);
-
-                if (stat.isDirectory()) {
-                    count += this.countFilesInDirectory(fullPath);
-                } else if (this.isTextFile(fullPath)) {
-                    count++;
-                }
-            }
-        } catch { }
-        return count;
+        const stat = fs.statSync(fullPath);
+        // Template/profile patterns select files; directories are always traversed
+        if (!stat.isDirectory() && !this.passesPatternFilters(relativePath)) {
+          continue;
+        }
+        if (stat.isDirectory()) {
+          if (!FileUtils.isSkippedDirectory(item)) {
+            files.push(...this.scanDirectory(fullPath));
+          }
+        } else if (this.isTextFile(fullPath) && this.passesFileFilter(relativePath)) {
+          // Denied by the policy: left out of the analysis and every export
+          if (this.policy?.isDenied(relativePath)) {
+            this.stats.policyDeniedFiles++;
+          } else {
+            files.push(fullPath);
+          }
+        }
+      }
+    } catch (error) {
+      console.error(`Error scanning directory ${dir}:`, error.message);
     }
 
-    updateStats(fileInfo) {
-        if (fileInfo.error) return;
-
-        const { extension, tokens, sizeBytes, lines } = fileInfo;
-
-        this.stats.totalFiles++;
-        this.stats.totalTokens += tokens;
-        this.stats.totalBytes += sizeBytes;
-        this.stats.totalLines += lines;
-        this.stats.largestFiles.push(fileInfo);
-
-        // Update extension stats
-        this.stats.byExtension[extension] = this.stats.byExtension[extension] ||
-            { count: 0, tokens: 0, bytes: 0, lines: 0 };
-        const extStats = this.stats.byExtension[extension];
-        extStats.count++; extStats.tokens += tokens;
-        extStats.bytes += sizeBytes; extStats.lines += lines;
-
-        // Update directory stats
-        const topDir = fileInfo.relativePath.split('/')[0];
-        this.stats.byDirectory[topDir] = this.stats.byDirectory[topDir] ||
-            { count: 0, tokens: 0, bytes: 0, lines: 0 };
-        const dirStats = this.stats.byDirectory[topDir];
-        dirStats.count++; dirStats.tokens += tokens;
-        dirStats.bytes += sizeBytes; dirStats.lines += lines;
+    return files;
+  }
+
+  /**
+   * Optional caller-supplied filter (e.g. --changed-only) on project-relative paths
+   * @param {string} relativePath
+   * @returns {boolean}
+   */
+  passesFileFilter(relativePath) {
+    return typeof this.options.fileFilter !== 'function' || this.options.fileFilter(relativePath);
+  }
+
+  /**
+   * @param {string} relativePath - Path relative to the project root
+   * @returns {boolean} True for a report, export, digest or state folder written by ctxman
+   */
+  isOwnOutput(relativePath) {
+    return (
+      OWN_OUTPUT_PATHS.some((pattern) => pattern.test(relativePath)) ||
+      Boolean(this.policy?.isAuditLog(relativePath))
+    );
+  }
+
+  /**
+   * @param {string} fullPath - File or directory matched by an ignore rule
+   * @returns {number} Files it stands for
+   */
+  countMatchedFiles(fullPath) {
+    try {
+      return fs.statSync(fullPath).isDirectory() ? this.countFilesInDirectory(fullPath) : 1;
+    } catch {
+      return 0;
     }
-
-    generateLLMContext(analysisResults) {
-        const context = {
-            project: {
-                root: path.basename(this.projectRoot),
-                totalFiles: this.stats.totalFiles,
-                totalTokens: this.stats.totalTokens
-            }
-        };
-
-        if (this.options.methodLevel) {
-            context.methods = this.generateMethodContext(analysisResults.files || analysisResults);
-            context.methodStats = {
-                totalMethods: this.methodStats.totalMethods,
-                includedMethods: this.methodStats.includedMethods,
-                totalMethodTokens: Object.values(this.methodStats.methodTokens).reduce((sum, tokens) => sum + tokens, 0)
-            };
+  }
+
+  countIgnoredFiles(filePath) {
+    // Reasons set by GitIgnoreParser for .contextignore / .contextinclude matches
+    const isCalculatorIgnored = ['context', 'context-include'].includes(
+      this.gitIgnore._lastIgnoreReason
+    );
+
+    try {
+      const stat = fs.statSync(filePath);
+      if (stat.isDirectory()) {
+        const count = this.countFilesInDirectory(filePath);
+        if (isCalculatorIgnored) {
+          this.stats.calculatorIgnoredFiles += count;
         } else {
-            context.paths = this.generateCompactPaths(analysisResults.files || analysisResults);
+          this.stats.ignoredFiles += count;
         }
-
-        return context;
-    }
-
-    generateMethodContext(analysisResults) {
-        const methodContext = {};
-
-        for (const fileInfo of analysisResults) {
-            if (fileInfo.methods && fileInfo.methods.length > 0) {
-                methodContext[fileInfo.relativePath] = fileInfo.methods.map(method => ({
-                    name: method.name,
-                    line: method.line,
-                    tokens: method.tokens
-                }));
-            }
+      } else {
+        if (isCalculatorIgnored) {
+          this.stats.calculatorIgnoredFiles++;
+        } else {
+          this.stats.ignoredFiles++;
         }
-
-        return methodContext;
+      }
+    } catch {}
+  }
+
+  countFilesInDirectory(dirPath) {
+    let count = 0;
+    try {
+      for (const item of fs.readdirSync(dirPath)) {
+        const fullPath = path.join(dirPath, item);
+        const stat = fs.statSync(fullPath);
+
+        if (stat.isDirectory()) {
+          count += this.countFilesInDirectory(fullPath);
+        } else if (this.isTextFile(fullPath)) {
+          count++;
+        }
+      }
+    } catch {}
+    return count;
+  }
+
+  updateStats(fileInfo) {
+    if (fileInfo.error) return;
+
+    const { extension, tokens, sizeBytes, lines } = fileInfo;
+
+    this.stats.totalFiles++;
+    this.stats.totalTokens += tokens;
+    this.stats.totalBytes += sizeBytes;
+    this.stats.totalLines += lines;
+    this.stats.largestFiles.push(fileInfo);
+
+    // Update extension stats
+    this.stats.byExtension[extension] = this.stats.byExtension[extension] || {
+      count: 0,
+      tokens: 0,
+      bytes: 0,
+      lines: 0,
+    };
+    const extStats = this.stats.byExtension[extension];
+    extStats.count++;
+    extStats.tokens += tokens;
+    extStats.bytes += sizeBytes;
+    extStats.lines += lines;
+
+    // Update directory stats
+    // Files in the project root share '.', instead of each counting as a directory
+    const parts = fileInfo.relativePath.split(/[\\/]/);
+    const topDir = parts.length > 1 ? parts[0] : '.';
+    this.stats.byDirectory[topDir] = this.stats.byDirectory[topDir] || {
+      count: 0,
+      tokens: 0,
+      bytes: 0,
+      lines: 0,
+    };
+    const dirStats = this.stats.byDirectory[topDir];
+    dirStats.count++;
+    dirStats.tokens += tokens;
+    dirStats.bytes += sizeBytes;
+    dirStats.lines += lines;
+  }
+
+  generateLLMContext(analysisResults) {
+    const context = {
+      project: {
+        root: path.basename(this.projectRoot),
+        totalFiles: this.stats.totalFiles,
+        totalTokens: this.stats.totalTokens,
+      },
+    };
+
+    if (this.options.methodLevel) {
+      context.methods = this.generateMethodContext(analysisResults.files || analysisResults);
+      context.methodStats = {
+        totalMethods: this.methodStats.totalMethods,
+        includedMethods: this.methodStats.includedMethods,
+        totalMethodTokens: Object.values(this.methodStats.methodTokens).reduce(
+          (sum, tokens) => sum + tokens,
+          0
+        ),
+      };
+    } else {
+      context.paths = this.generateCompactPaths(analysisResults.files || analysisResults);
     }
 
-    generateCompactPaths(analysisResults) {
-        const sortedFiles = analysisResults.sort((a, b) => b.tokens - a.tokens);
-        const pathGroups = {};
-
-        sortedFiles.forEach(file => {
-            const filePath = file.relativePath;
-            const parts = filePath.split('/');
-
-            // Group files by their directory path
-            // If file is in root, use '/' as key, otherwise use directory path with trailing slash
-            const categoryPath = parts.length > 1 ? parts.slice(0, -1).join('/') + '/' : '/';
+    return context;
+  }
 
-            if (!pathGroups[categoryPath]) {
-                pathGroups[categoryPath] = [];
-            }
+  generateMethodContext(analysisResults) {
+    const methodContext = {};
 
-            const filename = parts[parts.length - 1];
-            pathGroups[categoryPath].push(filename);
-        });
-
-        return pathGroups;
+    for (const fileInfo of analysisResults) {
+      if (fileInfo.methods && fileInfo.methods.length > 0) {
+        methodContext[fileInfo.relativePath] = fileInfo.methods.map((method) => ({
+          name: method.name,
+          line: method.line,
+          tokens: method.tokens,
+        }));
+      }
     }
 
-    exportContextToClipboard(context) {
-        const contextString = JSON.stringify(context, null, 2);
-        const success = ClipboardUtils.copy(contextString);
-
-        if (!success) {
-            console.log('💡 Context saved to llm-context.json instead');
-            this.saveContextToFile(context);
-        }
+    return methodContext;
+  }
+
+  generateCompactPaths(analysisResults) {
+    const sortedFiles = analysisResults.sort((a, b) => b.tokens - a.tokens);
+    const pathGroups = {};
+
+    sortedFiles.forEach((file) => {
+      const filePath = file.relativePath;
+      const parts = filePath.split('/');
+
+      // Group files by their directory path
+      // If file is in root, use '/' as key, otherwise use directory path with trailing slash
+      const categoryPath = parts.length > 1 ? parts.slice(0, -1).join('/') + '/' : '/';
+
+      if (!pathGroups[categoryPath]) {
+        pathGroups[categoryPath] = [];
+      }
+
+      const filename = parts[parts.length - 1];
+      pathGroups[categoryPath].push(filename);
+    });
+
+    return pathGroups;
+  }
+
+  /**
+   * Format chosen with -o/--output for context exports (JSON when not set)
+   * @returns {string}
+   */
+  getContextFormat() {
+    const format = this.options.outputFormat;
+    return format && format !== 'gitingest' ? format : 'json';
+  }
+
+  /**
+   * Serialize the LLM context in the selected output format
+   * @param {object} context
+   * @returns {string}
+   */
+  formatContext(context) {
+    const format = this.getContextFormat();
+    if (format === 'json') return JSON.stringify(context, null, 2);
+    return new FormatRegistry().encode(format, context);
+  }
+
+  exportContextToClipboard(context) {
+    const contextString = this.formatContext(context);
+    const success = ClipboardUtils.copy(contextString);
+
+    if (success) {
+      this.recordAudit('cli --context-clipboard', {
+        destination: 'clipboard',
+        format: this.getContextFormat(),
+        text: contextString,
+      });
+    } else {
+      console.log('💡 Clipboard unavailable, saving to a file instead');
+      this.saveContextToFile(context);
     }
-
-    saveContextToFile(context) {
-        const contextPath = path.join(this.projectRoot, 'llm-context.json');
-        fs.writeFileSync(contextPath, JSON.stringify(context, null, 2));
-        console.log(`💾 Context saved to: ${contextPath}`);
+  }
+
+  saveContextToFile(context) {
+    const format = this.getContextFormat();
+    const extension = format === 'json' ? '.json' : new FormatRegistry().getInfo(format).extension;
+    const contextPath = path.join(this.projectRoot, `llm-context${extension}`);
+    const text = this.formatContext(context);
+    fs.writeFileSync(contextPath, text);
+    console.log(`💾 Context saved to: ${contextPath}`);
+    this.recordAudit('cli --context-export', { destination: contextPath, format, text });
+  }
+
+  promptForExport(analysisResults) {
+    console.log('\n📤 Export Options:');
+    console.log('1) Save detailed JSON report (token-analysis-report.json)');
+    console.log('2) Generate LLM context file (llm-context.json)');
+    console.log('3) Copy LLM context to clipboard');
+    console.log('4) Generate GitIngest digest (digest.txt)');
+    console.log('5) No export (skip)');
+
+    const rl = readline.createInterface({
+      input: process.stdin,
+      output: process.stdout,
+    });
+
+    rl.question('🤔 Which export option would you like? (1-5): ', (answer) => {
+      const trimmed = answer.trim();
+
+      // Handle empty input (just Enter pressed) - default to option 5 (skip)
+      if (trimmed === '') {
+        console.log('\n✅ Analysis complete. No export selected.');
+        rl.close();
+        return;
+      }
+
+      const choice = parseInt(trimmed);
+
+      // Validate input is a number between 1-5
+      if (isNaN(choice) || choice < 1 || choice > 5) {
+        console.log('\n❌ Invalid option. Please enter a number between 1-5.');
+        console.log('   Skipping export.');
+        rl.close();
+        return;
+      }
+
+      const requested = ['--save-report', '--context-export', '--context-clipboard', '--gitingest'];
+      if (choice !== 5 && !this.checkExportPolicy(analysisResults, requested[choice - 1])) {
+        rl.close();
+        return;
+      }
+
+      switch (choice) {
+        case 1:
+          console.log('\n💾 Saving detailed JSON report...');
+          this.saveDetailedReport(analysisResults);
+          break;
+        case 2:
+          console.log('\n🤖 Generating LLM context file...');
+          const context = this.generateLLMContext(analysisResults);
+          this.saveContextToFile(context);
+          break;
+        case 3:
+          console.log('\n🤖 Generating LLM context for clipboard...');
+          const clipboardContext = this.generateLLMContext(analysisResults);
+          this.exportContextToClipboard(clipboardContext);
+          break;
+        case 4:
+          console.log('\n📄 Generating GitIngest digest...');
+          this.saveGitIngestDigest(analysisResults);
+          break;
+        case 5:
+          console.log('\n✅ Analysis complete. No export selected.');
+          break;
+      }
+
+      rl.close();
+    });
+  }
+
+  saveGitIngestDigest(analysisResults) {
+    const formatter = new GitIngestFormatter(this.projectRoot, this.stats, analysisResults, {
+      redactSecrets: this.options.redactSecrets !== false,
+      chunking: this.options.chunking,
+    });
+    const digestPath = path.join(this.projectRoot, 'digest.txt');
+    const digestSize = formatter.saveToFile(digestPath);
+
+    if (formatter.chunkFiles?.length) {
+      console.log(`💾 GitIngest digest saved as ${formatter.chunkFiles.length} chunks:`);
+      formatter.chunkFiles.forEach((file) => console.log(`   ${path.basename(file)}`));
+    } else {
+      console.log(`💾 GitIngest digest saved to: digest.txt`);
     }
-
-    promptForExport(analysisResults) {
-        console.log('\n📤 Export Options:');
-        console.log('1) Save detailed JSON report (token-analysis-report.json)');
-        console.log('2) Generate LLM context file (llm-context.json)');
-        console.log('3) Copy LLM context to clipboard');
-        console.log('4) Generate GitIngest digest (digest.txt)');
-        console.log('5) No export (skip)');
-
-        const rl = readline.createInterface({
-            input: process.stdin,
-            output: process.stdout
-        });
-
-        rl.question('🤔 Which export option would you like? (1-5): ', (answer) => {
-            const trimmed = answer.trim();
-
-            // Handle empty input (just Enter pressed) - default to option 5 (skip)
-            if (trimmed === '') {
-                console.log('\n✅ Analysis complete. No export selected.');
-                rl.close();
-                return;
-            }
-
-            const choice = parseInt(trimmed);
-
-            // Validate input is a number between 1-5
-            if (isNaN(choice) || choice < 1 || choice > 5) {
-                console.log('\n❌ Invalid option. Please enter a number between 1-5.');
-                console.log('   Skipping export.');
-                rl.close();
-                return;
-            }
-
-            switch (choice) {
-                case 1:
-                    console.log('\n💾 Saving detailed JSON report...');
-                    this.saveDetailedReport(analysisResults);
-                    break;
-                case 2:
-                    console.log('\n🤖 Generating LLM context file...');
-                    const context = this.generateLLMContext(analysisResults);
-                    this.saveContextToFile(context);
-                    console.log('💾 LLM context saved to: llm-context.json');
-                    break;
-                case 3:
-                    console.log('\n🤖 Generating LLM context for clipboard...');
-                    const clipboardContext = this.generateLLMContext(analysisResults);
-                    this.exportContextToClipboard(clipboardContext);
-                    break;
-                case 4:
-                    console.log('\n📄 Generating GitIngest digest...');
-                    this.saveGitIngestDigest(analysisResults);
-                    break;
-                case 5:
-                    console.log('\n✅ Analysis complete. No export selected.');
-                    break;
-            }
-
-            rl.close();
-        });
+    console.log(`📊 Digest size: ${(digestSize / 1024).toFixed(1)} KB`);
+    if (this.policy?.audit.enabled) {
+      const written = formatter.chunkFiles?.length ? formatter.chunkFiles : [digestPath];
+      this.recordAudit('cli --gitingest', {
+        destination: digestPath,
+        outputFiles: written.length > 1 ? written : undefined,
+        format: 'gitingest',
+        redactions: formatter.redactions,
+        // Hash what was written: the chunks in order
+        text: written.map((file) => fs.readFileSync(file)),
+      });
     }
-
-    saveGitIngestDigest(analysisResults) {
-        const formatter = new GitIngestFormatter(this.projectRoot, this.stats, analysisResults);
-        const digestPath = path.join(this.projectRoot, 'digest.txt');
-        const digestSize = formatter.saveToFile(digestPath);
-
-        console.log(`💾 GitIngest digest saved to: digest.txt`);
-        console.log(`📊 Digest size: ${(digestSize / 1024).toFixed(1)} KB`);
+    const redactions = formatter.redactions ?? {};
+    if (Object.keys(redactions).length > 0) {
+      console.log(
+        `🔒 Redacted secrets: ${formatRedactionSummary(redactions)} (disable with --no-redact)`
+      );
+    }
+  }
+
+  /**
+   * Scan and analyze the project without printing or exporting
+   * @returns {Array<object>} Per-file analysis results
+   */
+  analyze() {
+    return this.analyzeFiles(this.scanDirectory(this.projectRoot));
+  }
+
+  analyzeFiles(files) {
+    const analysisResults = [];
+    for (const file of files) {
+      const fileInfo = this.analyzeFile(file);
+      this.updateStats(fileInfo);
+      analysisResults.push(fileInfo);
     }
 
-    run() {
-        this.printHeader();
-
-        const allFiles = this.scanDirectory(this.projectRoot);
-        this.printScanResults(allFiles);
-
-        // Analyze all files
-        const analysisResults = [];
-        for (const file of allFiles) {
-            const fileInfo = this.analyzeFile(file);
-            this.updateStats(fileInfo);
-            analysisResults.push(fileInfo);
-        }
+    this.stats.largestFiles.sort((a, b) => b.tokens - a.tokens);
+    this.lastAnalysisResults = analysisResults;
+    return analysisResults;
+  }
 
-        this.stats.largestFiles.sort((a, b) => b.tokens - a.tokens);
-        this.printReport();
+  run() {
+    this.printHeader();
 
-        // Context Fit Analysis (v2.3.7)
-        if (this.options.targetModel && !this.options.dashboard) {
-            this.printContextFitAnalysis();
-        }
+    const allFiles = this.scanDirectory(this.projectRoot);
+    this.printScanResults(allFiles);
 
-        // Handle exports (skip for dashboard mode)
-        if (!this.options.dashboard) {
-            this.handleExports(analysisResults);
-        }
+    const analysisResults = this.analyzeFiles(allFiles);
+    this.printReport();
 
-        // Return stats for programmatic usage (e.g., Dashboard)
-        return this.stats;
+    // Context Fit Analysis (v2.3.7)
+    if (this.options.targetModel && !this.options.dashboard) {
+      this.printContextFitAnalysis();
     }
 
-    printHeader() {
-        console.log('🔍 Analyzing project: ' + this.projectRoot);
-        console.log('📝 Respecting .gitignore rules...');
-
-        const configPath = this.gitIgnore.hasIncludeFile ?
-            ConfigUtils.findConfigFile(this.projectRoot, '.contextinclude') :
-            ConfigUtils.findConfigFile(this.projectRoot, '.contextignore');
+    // Handle exports (skip for dashboard mode)
+    if (!this.options.dashboard) {
+      this.handleExports(analysisResults);
+    }
 
-        if (configPath) {
-            const mode = this.gitIgnore.hasIncludeFile ? 'INCLUDE' : 'EXCLUDE';
-            console.log(`📅 Found calculator config - using ${mode} mode`);
-            console.log('   Location: ' + path.relative(this.projectRoot, configPath));
-        }
+    // Return stats for programmatic usage (e.g., Dashboard)
+    return this.stats;
+  }
 
-        console.log('🎯 Token calculation: ' + (TokenUtils.hasExactCounting() ? '✅ Exact (using tiktoken)' : '⚠️  Estimated'));
+  printHeader() {
+    console.log('🔍 Analyzing project: ' + this.projectRoot);
+    console.log('📝 Respecting .gitignore rules...');
 
-        if (this.options.methodLevel) {
-            console.log('🔧 Method-level analysis: ✅ Enabled');
-            if (this.methodFilter) {
-                const hasInclude = this.methodFilter.hasIncludeFile;
-                const mode = hasInclude ? 'INCLUDE' : 'EXCLUDE';
-                console.log(`   Method filtering: ${mode} mode active`);
-            }
-        }
+    const configPath = this.gitIgnore.hasIncludeFile
+      ? ConfigUtils.findConfigFile(this.projectRoot, '.contextinclude')
+      : ConfigUtils.findConfigFile(this.projectRoot, '.contextignore');
 
-        console.log();
+    if (configPath) {
+      const mode = this.gitIgnore.hasIncludeFile ? 'INCLUDE' : 'EXCLUDE';
+      console.log(`📅 Found calculator config - using ${mode} mode`);
+      console.log('   Location: ' + path.relative(this.projectRoot, configPath));
     }
 
-    printScanResults(allFiles) {
-        console.log(`📁 Found ${allFiles.length} text files to analyze`);
-        console.log(`🚫 Ignored ${this.stats.ignoredFiles} files due to .gitignore rules`);
-        if (this.stats.calculatorIgnoredFiles > 0) {
-            const mode = this.gitIgnore.hasIncludeFile ? 'include rules' : 'ignore rules';
-            console.log(`📋 Filtered ${this.stats.calculatorIgnoredFiles} additional files due to calculator ${mode}`);
-        }
-
-        if (this.options.methodLevel) {
-            console.log(`🔧 Total methods found: ${this.methodStats.totalMethods}`);
-            console.log(`✅ Methods included: ${this.methodStats.includedMethods}`);
-        }
-
-        console.log();
+    if (this.policy) {
+      const files = this.policy.sources.map(({ scope, file }) =>
+        scope === 'project' ? POLICY_FILE : file
+      );
+      console.log(`🛡️  Export policy: ${files.join(', ')}`);
+    }
+    if (this.ignoreFileRules) {
+      console.log(`🧩 AI-tool ignore files: ${this.ignoreFileRules.describe()}`);
     }
 
-    printReport() {
-        console.log('='.repeat(80));
-        console.log('🎯 PROJECT TOKEN ANALYSIS REPORT');
-        console.log('='.repeat(80));
-        console.log(`📊 Total files analyzed: ${this.stats.totalFiles.toLocaleString()}`);
-        console.log(`🔢 Total tokens: ${this.stats.totalTokens.toLocaleString()}`);
-        console.log(`💾 Total size: ${(this.stats.totalBytes / 1024 / 1024).toFixed(2)} MB`);
-        console.log(`📄 Total lines: ${this.stats.totalLines.toLocaleString()}`);
-
-        const avgTokens = this.stats.totalFiles > 0
-            ? Math.round(this.stats.totalTokens / this.stats.totalFiles).toLocaleString()
-            : 'N/A';
-        console.log(`📈 Average tokens per file: ${avgTokens}`);
-        console.log(`🚫 Files ignored by .gitignore: ${this.stats.ignoredFiles.toLocaleString()}`);
-        if (this.stats.calculatorIgnoredFiles > 0) {
-            console.log(`📋 Files ignored by calculator rules: ${this.stats.calculatorIgnoredFiles.toLocaleString()}`);
-        }
-        console.log();
+    console.log(
+      '🎯 Token calculation: ' +
+        (TokenUtils.hasExactCounting()
+          ? `✅ Exact (using tiktoken) · ${this.encoding}`
+          : '⚠️  Estimated')
+    );
+    const target = this.options.targetModel
+      ? LLMDetector.getProfile(this.options.targetModel)
+      : null;
+    if (target?.tokenizer && !target.tokenizer.exact && TokenUtils.hasExactCounting()) {
+      const exactly =
+        target.tokenizer.exactVia === 'anthropic-api'
+          ? ` (exact count: ctxman count --model ${target.id} --api)`
+          : '';
+      console.log(
+        `   Counts for ${target.name} are approximate: it uses its own tokenizer${exactly}`
+      );
+    }
 
-        this.printExtensionStats();
-        this.printTopFiles();
-        this.printTopDirectories();
+    if (this.options.methodLevel) {
+      console.log('🔧 Method-level analysis: ✅ Enabled');
+      if (this.methodFilter) {
+        const hasInclude = this.methodFilter.hasIncludeFile;
+        const mode = hasInclude ? 'INCLUDE' : 'EXCLUDE';
+        console.log(`   Method filtering: ${mode} mode active`);
+      }
+    }
 
-        console.log('\n💡 Tip: Use --save-report flag to save detailed analysis to JSON file');
+    console.log();
+  }
+
+  printScanResults(allFiles) {
+    console.log(`📁 Found ${allFiles.length} text files to analyze`);
+    console.log(`🚫 Ignored ${this.stats.ignoredFiles} files due to .gitignore rules`);
+    if (this.stats.calculatorIgnoredFiles > 0) {
+      const rules = this.gitIgnore.hasIncludeFile ? '.contextinclude' : '.contextignore';
+      console.log(
+        `📋 Filtered ${this.stats.calculatorIgnoredFiles} additional files due to ${rules} rules`
+      );
+    }
+    if (this.stats.ignoreFileIgnoredFiles > 0) {
+      console.log(
+        `🧩 Filtered ${this.stats.ignoreFileIgnoredFiles} files due to AI-tool ignore files`
+      );
+    }
+    if (this.stats.policyDeniedFiles > 0) {
+      const count = this.stats.policyDeniedFiles;
+      console.log(
+        `🛡️  Removed ${count} file${count === 1 ? '' : 's'} denied by the export policy (never exported)`
+      );
     }
 
-    printExtensionStats() {
-        console.log('📋 BY FILE TYPE:');
-        console.log('-'.repeat(80));
-        console.log('Extension'.padEnd(15) + 'Files'.padStart(8) + 'Tokens'.padStart(12) + 'Size (KB)'.padStart(12) + 'Lines'.padStart(10));
-        console.log('-'.repeat(80));
-
-        Object.entries(this.stats.byExtension)
-            .sort((a, b) => b[1].tokens - a[1].tokens)
-            .forEach(([ext, data]) => {
-                console.log(
-                    ext.padEnd(15) +
-                    data.count.toString().padStart(8) +
-                    data.tokens.toLocaleString().padStart(12) +
-                    (data.bytes / 1024).toFixed(1).padStart(12) +
-                    data.lines.toLocaleString().padStart(10)
-                );
-            });
+    if (this.options.methodLevel) {
+      console.log(`🔧 Total methods found: ${this.methodStats.totalMethods}`);
+      console.log(`✅ Methods included: ${this.methodStats.includedMethods}`);
     }
 
-    printTopFiles() {
-        console.log('\n🏆 TOP 5 LARGEST FILES BY TOKEN COUNT:');
-        console.log('-'.repeat(80));
-        this.stats.largestFiles.slice(0, 5).forEach((file, index) => {
-            const percentage = ((file.tokens / this.stats.totalTokens) * 100).toFixed(1);
-            const sizeKB = (file.sizeBytes / 1024).toFixed(1);
-            console.log(`${(index + 1).toString().padStart(2)}. ${file.tokens.toLocaleString().padStart(8)} tokens (${percentage}%) - ${file.relativePath}`);
-            console.log(`    💾 ${sizeKB.padStart(8)} KB, 📄 ${file.lines.toLocaleString().padStart(6)} lines`);
-        });
+    console.log();
+  }
+
+  printReport() {
+    console.log('='.repeat(80));
+    console.log('🎯 PROJECT TOKEN ANALYSIS REPORT');
+    console.log('='.repeat(80));
+    console.log(`📊 Total files analyzed: ${this.stats.totalFiles.toLocaleString()}`);
+    console.log(`🔢 Total tokens: ${this.stats.totalTokens.toLocaleString()}`);
+    console.log(`💾 Total size: ${(this.stats.totalBytes / 1024 / 1024).toFixed(2)} MB`);
+    console.log(`📄 Total lines: ${this.stats.totalLines.toLocaleString()}`);
+
+    const avgTokens =
+      this.stats.totalFiles > 0
+        ? Math.round(this.stats.totalTokens / this.stats.totalFiles).toLocaleString()
+        : 'N/A';
+    console.log(`📈 Average tokens per file: ${avgTokens}`);
+    console.log(`🚫 Files ignored by .gitignore: ${this.stats.ignoredFiles.toLocaleString()}`);
+    if (this.stats.calculatorIgnoredFiles > 0) {
+      console.log(
+        `📋 Files ignored by context rules: ${this.stats.calculatorIgnoredFiles.toLocaleString()}`
+      );
     }
+    console.log();
 
-    printTopDirectories() {
-        const largestDirectories = Object.entries(this.stats.byDirectory)
-            .map(([dir, stats]) => ({ directory: dir, ...stats }))
-            .sort((a, b) => b.tokens - a.tokens);
-
-        console.log('\n📁 TOP 5 LARGEST DIRECTORIES BY TOKEN COUNT:');
-        console.log('-'.repeat(80));
-        largestDirectories.slice(0, 5).forEach((dir, index) => {
-            const percentage = ((dir.tokens / this.stats.totalTokens) * 100).toFixed(1);
-            const avgTokensPerFile = Math.round(dir.tokens / dir.count);
-            console.log(`${(index + 1).toString().padStart(2)}. ${dir.tokens.toLocaleString().padStart(8)} tokens (${percentage}%) - ${dir.directory}/`);
-            console.log(`    📄 ${dir.count.toString().padStart(3)} files, 📈 ${avgTokensPerFile.toLocaleString().padStart(6)} avg tokens/file`);
-        });
+    this.printExtensionStats();
+    this.printTopFiles();
+    this.printTopDirectories();
+    if (this.options.verbose) {
+      this.printIncludedFiles();
     }
 
-    handleExports(analysisResults) {
-        const { saveReport, contextExport, contextToClipboard, gitingest } = this.options;
-
-        if (contextExport || contextToClipboard || saveReport || gitingest) {
-            if (contextExport || contextToClipboard) {
-                console.log('\n🤖 Generating LLM Context...');
-                const context = this.generateLLMContext(analysisResults);
-
-                if (contextToClipboard) {
-                    this.exportContextToClipboard(context);
-                } else {
-                    this.saveContextToFile(context);
-                    console.log('💾 LLM context saved to: llm-context.json');
-                }
-            }
-
-            if (saveReport) {
-                this.saveDetailedReport(analysisResults);
-            }
-
-            if (gitingest) {
-                console.log('\n📄 Generating GitIngest digest...');
-                this.saveGitIngestDigest(analysisResults);
-            }
+    console.log('\n💡 Tip: Use --save-report flag to save detailed analysis to JSON file');
+  }
+
+  printExtensionStats() {
+    console.log('📋 BY FILE TYPE:');
+    console.log('-'.repeat(80));
+    console.log(
+      'Extension'.padEnd(15) +
+        'Files'.padStart(8) +
+        'Tokens'.padStart(12) +
+        'Size (KB)'.padStart(12) +
+        'Lines'.padStart(10)
+    );
+    console.log('-'.repeat(80));
+
+    Object.entries(this.stats.byExtension)
+      .sort((a, b) => b[1].tokens - a[1].tokens)
+      .forEach(([ext, data]) => {
+        console.log(
+          ext.padEnd(15) +
+            data.count.toString().padStart(8) +
+            data.tokens.toLocaleString().padStart(12) +
+            (data.bytes / 1024).toFixed(1).padStart(12) +
+            data.lines.toLocaleString().padStart(10)
+        );
+      });
+  }
+
+  printTopFiles() {
+    console.log('\n🏆 TOP 5 LARGEST FILES BY TOKEN COUNT:');
+    console.log('-'.repeat(80));
+    this.stats.largestFiles.slice(0, 5).forEach((file, index) => {
+      const percentage = ((file.tokens / this.stats.totalTokens) * 100).toFixed(1);
+      const sizeKB = (file.sizeBytes / 1024).toFixed(1);
+      console.log(
+        `${(index + 1).toString().padStart(2)}. ${file.tokens.toLocaleString().padStart(8)} tokens (${percentage}%) - ${file.relativePath}`
+      );
+      console.log(
+        `    💾 ${sizeKB.padStart(8)} KB, 📄 ${file.lines.toLocaleString().padStart(6)} lines`
+      );
+    });
+  }
+
+  printTopDirectories() {
+    const largestDirectories = Object.entries(this.stats.byDirectory)
+      .map(([dir, stats]) => ({ directory: dir, ...stats }))
+      .sort((a, b) => b.tokens - a.tokens);
+
+    console.log('\n📁 TOP 5 LARGEST DIRECTORIES BY TOKEN COUNT:');
+    console.log('-'.repeat(80));
+    largestDirectories.slice(0, 5).forEach((dir, index) => {
+      const percentage = ((dir.tokens / this.stats.totalTokens) * 100).toFixed(1);
+      const avgTokensPerFile = Math.round(dir.tokens / dir.count);
+      console.log(
+        `${(index + 1).toString().padStart(2)}. ${dir.tokens.toLocaleString().padStart(8)} tokens (${percentage}%) - ${dir.directory}/`
+      );
+      console.log(
+        `    📄 ${dir.count.toString().padStart(3)} files, 📈 ${avgTokensPerFile.toLocaleString().padStart(6)} avg tokens/file`
+      );
+    });
+  }
+
+  printIncludedFiles() {
+    console.log(`\n📄 INCLUDED FILES (${this.stats.largestFiles.length}):`);
+    console.log('-'.repeat(80));
+    [...this.stats.largestFiles]
+      .sort((a, b) => a.relativePath.localeCompare(b.relativePath))
+      .forEach((file) => {
+        console.log(`${file.tokens.toLocaleString().padStart(8)} tokens - ${file.relativePath}`);
+      });
+  }
+
+  handleExports(analysisResults) {
+    const { saveReport, contextExport, contextToClipboard } = this.options;
+    // -o gitingest and --chunk both mean "produce the GitIngest digest"
+    const gitingest =
+      this.options.gitingest ||
+      this.options.outputFormat === 'gitingest' ||
+      Boolean(this.options.chunking?.enabled);
+
+    if (contextExport || contextToClipboard || saveReport || gitingest) {
+      const requested = [
+        contextToClipboard ? '--context-clipboard' : contextExport && '--context-export',
+        saveReport && '--save-report',
+        gitingest && '--gitingest',
+      ].filter(Boolean);
+      if (!this.checkExportPolicy(analysisResults, requested.join(' '))) return;
+
+      if (contextExport || contextToClipboard) {
+        console.log('\n🤖 Generating LLM Context...');
+        const context = this.generateLLMContext(analysisResults);
+
+        if (contextToClipboard) {
+          this.exportContextToClipboard(context);
         } else {
-            this.promptForExport(analysisResults);
+          this.saveContextToFile(context);
         }
+      }
+
+      if (saveReport) {
+        this.saveDetailedReport(analysisResults);
+      }
+
+      if (gitingest) {
+        console.log('\n📄 Generating GitIngest digest...');
+        this.saveGitIngestDigest(analysisResults);
+      }
+    } else if (!this.options.simple) {
+      this.promptForExport(analysisResults);
     }
-
-    saveDetailedReport(analysisResults) {
-        const report = {
-            metadata: {
-                generatedAt: new Date().toISOString(),
-                projectRoot: this.projectRoot,
-                gitignoreRules: this.gitIgnore.patterns.map(p => p.original),
-                calculatorRules: this.gitIgnore.contextPatterns.map(p => p.original)
-            },
-            summary: this.stats,
-            files: analysisResults.sort((a, b) => b.tokens - a.tokens)
-        };
-
-        const reportPath = path.join(this.projectRoot, 'token-analysis-report.json');
-        fs.writeFileSync(reportPath, JSON.stringify(report, null, 2));
-        console.log('💾 Detailed analysis saved to: token-analysis-report.json');
+  }
+
+  saveDetailedReport(analysisResults) {
+    const files = analysisResults.sort((a, b) => b.tokens - a.tokens);
+    // Method-level reports carry method source; a policy that requires redaction masks it
+    const redactor = this.policy?.requireRedaction ? this.createMethodRedactor() : null;
+    const report = {
+      metadata: {
+        generatedAt: new Date().toISOString(),
+        projectRoot: this.projectRoot,
+        gitignoreRules: this.gitIgnore.patterns.map((p) => p.original),
+        calculatorRules: this.gitIgnore.contextPatterns.map((p) => p.original),
+      },
+      summary: redactor
+        ? { ...this.stats, largestFiles: this.stats.largestFiles.map(redactor.file) }
+        : this.stats,
+      files: redactor ? files.map(redactor.file) : files,
+    };
+
+    const reportPath = path.join(this.projectRoot, 'token-analysis-report.json');
+    const text = JSON.stringify(report, null, 2);
+    fs.writeFileSync(reportPath, text);
+    console.log('💾 Detailed analysis saved to: token-analysis-report.json');
+    this.recordAudit('cli --save-report', {
+      destination: reportPath,
+      format: 'json-report',
+      redactions: redactor?.count() ?? 0,
+      text,
+    });
+  }
+
+  /**
+   * Masks secrets in the method source of analysis results. A result listed twice in the
+   * report (files and summary.largestFiles) is copied and counted once.
+   * @returns {{file: (entry: object) => object, count: () => number}}
+   */
+  createMethodRedactor() {
+    const copies = new Map();
+    let count = 0;
+    const redactMethod = (method) => {
+      if (typeof method?.content !== 'string') return method;
+      const { text, findings } = redactSecrets(method.content);
+      count += findings.length;
+      return { ...method, content: text };
+    };
+    const file = (entry) => {
+      if (!Array.isArray(entry?.methods)) return entry;
+      if (!copies.has(entry)) {
+        copies.set(entry, { ...entry, methods: entry.methods.map(redactMethod) });
+      }
+      return copies.get(entry);
+    };
+    return { file, count: () => count };
+  }
+
+  /**
+   * Check the requested exports against the export policy. A refused export is printed,
+   * logged with outcome "refused" and sets exit code 3.
+   * @param {Array<object>} analysisResults
+   * @param {string} [requested] - The export flags, for the audit log
+   * @returns {boolean} True when the exports may go ahead
+   */
+  checkExportPolicy(analysisResults, requested = 'export') {
+    if (!this.policy) return true;
+
+    try {
+      enforceExport(this.policy, {
+        files: analysisResults,
+        tokens: this.stats.totalTokens,
+        redact: this.options.redactSecrets !== false,
+        targetModel: this.options.targetModel || null,
+      });
+      return true;
+    } catch (error) {
+      if (!(error instanceof PolicyViolationError)) throw error;
+      console.error(`\n❌ ${error.message}`);
+      recordExport(this.policy, {
+        ...this.auditEvent(`cli ${requested}`),
+        outcome: 'refused',
+        violations: error.violations,
+      });
+      process.exitCode = error.exitCode;
+      return false;
     }
-
-    /**
-     * Print Context Fit Analysis (v2.3.7)
-     * Shows if repository fits in target LLM's context window
-     */
-    printContextFitAnalysis() {
-        try {
-            const profile = LLMDetector.getProfile(this.options.targetModel);
-            const repoStats = {
-                totalTokens: this.stats.totalTokens,
-                totalFiles: this.stats.totalFiles
-            };
-
-            const analysis = LLMDetector.analyzeContextFit(profile, repoStats);
-            const formatted = LLMDetector.formatAnalysis(analysis);
-
-            console.log(formatted);
-
-        } catch (error) {
-            // Silently fail if context fit analysis fails
-            // Don't break the main workflow
-        }
+  }
+
+  /**
+   * Audit event for an export of the current analysis
+   * @param {string} command
+   * @param {object} [extra] - destination, format, text, redactions, ...
+   * @returns {import('../policy/audit.js').AuditEvent}
+   */
+  auditEvent(command, extra = {}) {
+    return {
+      command,
+      root: this.projectRoot,
+      targetModel: this.options.targetModel || null,
+      files: this.lastAnalysisResults ?? [],
+      tokens: this.stats.totalTokens,
+      removedByPolicy: this.stats.policyDeniedFiles ?? 0,
+      ...extra,
+    };
+  }
+
+  /**
+   * Append an export to the policy's audit log (no-op without a policy that enables it)
+   * @param {string} command
+   * @param {object} extra
+   */
+  recordAudit(command, extra) {
+    if (this.policy?.audit.enabled) recordExport(this.policy, this.auditEvent(command, extra));
+  }
+
+  /**
+   * Print Context Fit Analysis (v2.3.7)
+   * Shows if repository fits in target LLM's context window
+   */
+  printContextFitAnalysis() {
+    try {
+      const profile = LLMDetector.getProfile(this.options.targetModel);
+      const repoStats = {
+        totalTokens: this.stats.totalTokens,
+        totalFiles: this.stats.totalFiles,
+      };
+
+      // Pass largest files for suggestions
+      const analysis = LLMDetector.analyzeContextFit(profile, repoStats, this.stats.largestFiles);
+      const formatted = LLMDetector.formatAnalysis(analysis);
+
+      console.log(formatted);
+    } catch (_error) {
+      // Silently fail if context fit analysis fails
+      // Don't break the main workflow
     }
+  }
 }
 
-
-
-
 export default TokenCalculator;