@nerviq/cli 1.17.3 → 1.19.0

package/src/harmony/drift.js

@@ -1,401 +1,401 @@
-/**
- * Harmony Drift Detection Engine
- *
- * Compares configurations ACROSS platforms and finds inconsistencies
- * in instructions, trust posture, MCP servers, rules, and coverage.
- *
- * Drift types:
- *   instruction-drift  — different platforms say different things
- *   trust-drift        — sandbox/approval modes differ across platforms
- *   mcp-drift          — MCP servers not aligned across platforms
- *   rule-drift         — rule coverage differs between platforms
- *   coverage-gap       — platform missing instructions entirely
- *
- * Severity: critical, high, medium, low
- */
-
-const SEVERITY_ORDER = { critical: 3, high: 2, medium: 1, low: 0 };
-
-const COLORS = {
-  reset: '\x1b[0m',
-  bold: '\x1b[1m',
-  dim: '\x1b[2m',
-  red: '\x1b[31m',
-  green: '\x1b[32m',
-  yellow: '\x1b[33m',
-  blue: '\x1b[36m',
-  magenta: '\x1b[35m',
-};
-
-function colorize(text, color) {
-  return `${COLORS[color] || ''}${text}${COLORS.reset}`;
-}
-
-// ─── Trust level risk mapping ───────────────────────────────────────────────
-
-const TRUST_RISK = {
-  'bypass': 4,
-  'full-auto': 4,
-  'unrestricted': 4,
-  'no-sandbox': 3,
-  'standard': 2,
-  'safe-write': 2,
-  'default': 1,
-  'locked-down': 0,
-  'unknown': 1,
-};
-
-// ─── Drift detectors ───────────────────────────────────────────────────────
-
-/**
- * Detect trust posture drift across platforms.
- * Critical if risk gap >= 2 levels between any two platforms.
- */
-function detectTrustDrift(model) {
-  const drifts = [];
-  const platforms = Object.keys(model.trustPosture);
-
-  if (platforms.length < 2) return drifts;
-
-  const riskLevels = {};
-  for (const p of platforms) {
-    riskLevels[p] = TRUST_RISK[model.trustPosture[p]] ?? 1;
-  }
-
-  const maxRisk = Math.max(...Object.values(riskLevels));
-  const minRisk = Math.min(...Object.values(riskLevels));
-  const gap = maxRisk - minRisk;
-
-  if (gap === 0) return drifts;
-
-  const highRiskPlatforms = platforms.filter(p => riskLevels[p] === maxRisk);
-  const lowRiskPlatforms = platforms.filter(p => riskLevels[p] === minRisk);
-
-  let severity = 'low';
-  if (gap >= 3) severity = 'critical';
-  else if (gap >= 2) severity = 'high';
-  else if (gap >= 1) severity = 'medium';
-
-  drifts.push({
-    type: 'trust-drift',
-    severity,
-    platforms,
-    description: `Trust posture gap of ${gap} levels: ` +
-      `${highRiskPlatforms.map(p => `${p}=${model.trustPosture[p]}`).join(', ')} vs ` +
-      `${lowRiskPlatforms.map(p => `${p}=${model.trustPosture[p]}`).join(', ')}`,
-    recommendation: `Align trust posture across platforms. Consider raising ${lowRiskPlatforms.join(', ')} ` +
-      `or tightening ${highRiskPlatforms.join(', ')} to reduce risk surface.`,
-  });
-
-  return drifts;
-}
-
-/**
- * Detect MCP server drift — servers present on some platforms but not others.
- */
-function detectMcpDrift(model) {
-  const drifts = [];
-  const allPlatforms = model.activePlatforms.map(p => p.platform);
-  const mcpCapablePlatforms = allPlatforms.filter(p =>
-    p === 'claude' || p === 'gemini' || p === 'copilot' || p === 'cursor' || p === 'windsurf' || p === 'opencode'
-  );
-
-  if (mcpCapablePlatforms.length < 2) return drifts;
-
-  for (const [name, server] of Object.entries(model.mcpServers)) {
-    const present = server.platforms.filter(p => mcpCapablePlatforms.includes(p));
-    const missing = mcpCapablePlatforms.filter(p => !server.platforms.includes(p));
-
-    if (missing.length > 0 && present.length > 0) {
-      drifts.push({
-        type: 'mcp-drift',
-        severity: missing.length >= 2 ? 'high' : 'medium',
-        platforms: [...present, ...missing],
-        description: `MCP server "${name}" is configured on ${present.join(', ')} but missing from ${missing.join(', ')}`,
-        recommendation: `Add "${name}" to ${missing.join(', ')} MCP configuration to ensure consistent tool access.`,
-      });
-    }
-  }
-
-  // Check for platforms with zero MCP servers when others have some
-  const serverCounts = {};
-  for (const p of mcpCapablePlatforms) {
-    const detail = model.platformDetails[p];
-    serverCounts[p] = detail ? detail.mcpServers.length : 0;
-  }
-
-  const withServers = mcpCapablePlatforms.filter(p => serverCounts[p] > 0);
-  const withoutServers = mcpCapablePlatforms.filter(p => serverCounts[p] === 0);
-
-  if (withServers.length > 0 && withoutServers.length > 0) {
-    drifts.push({
-      type: 'mcp-drift',
-      severity: 'medium',
-      platforms: [...withServers, ...withoutServers],
-      description: `MCP servers configured on ${withServers.join(', ')} (${withServers.map(p => serverCounts[p]).join(', ')} servers) ` +
-        `but ${withoutServers.join(', ')} have none`,
-      recommendation: `Consider adding MCP servers to ${withoutServers.join(', ')} for consistent tool access.`,
-    });
-  }
-
-  return drifts;
-}
-
-/**
- * Detect rule drift — platforms with rules configured vs those without.
- */
-function detectRuleDrift(model) {
-  const drifts = [];
-  const allPlatforms = model.activePlatforms.map(p => p.platform);
-
-  // Platforms that support rules: claude (.claude/rules), copilot (.github/instructions), cursor (.cursor/rules)
-  const ruleCapable = allPlatforms.filter(p =>
-    p === 'claude' || p === 'copilot' || p === 'cursor' || p === 'windsurf'
-  );
-
-  if (ruleCapable.length < 2) return drifts;
-
-  const ruleCounts = {};
-  for (const p of ruleCapable) {
-    ruleCounts[p] = model.activePlatforms.find(ap => ap.platform === p)?.ruleCount || 0;
-  }
-
-  const withRules = ruleCapable.filter(p => ruleCounts[p] > 0);
-  const withoutRules = ruleCapable.filter(p => ruleCounts[p] === 0);
-
-  if (withRules.length > 0 && withoutRules.length > 0) {
-    drifts.push({
-      type: 'rule-drift',
-      severity: 'medium',
-      platforms: [...withRules, ...withoutRules],
-      description: `Rules configured on ${withRules.map(p => `${p} (${ruleCounts[p]})`).join(', ')} ` +
-        `but missing from ${withoutRules.join(', ')}`,
-      recommendation: `Add equivalent rules to ${withoutRules.join(', ')} to maintain consistent behavior constraints.`,
-    });
-  }
-
-  // Check for large rule count gaps
-  const maxRules = Math.max(...Object.values(ruleCounts));
-  const minRules = Math.min(...Object.values(ruleCounts));
-
-  if (maxRules > 0 && minRules > 0 && maxRules > minRules * 3) {
-    const richPlatform = ruleCapable.find(p => ruleCounts[p] === maxRules);
-    const sparPlatform = ruleCapable.find(p => ruleCounts[p] === minRules);
-    drifts.push({
-      type: 'rule-drift',
-      severity: 'low',
-      platforms: [richPlatform, sparPlatform],
-      description: `Large rule count gap: ${richPlatform} has ${maxRules} rules vs ${sparPlatform} with ${minRules}`,
-      recommendation: `Review whether ${sparPlatform} is missing important rules that ${richPlatform} enforces.`,
-    });
-  }
-
-  return drifts;
-}
-
-/**
- * Detect instruction coverage gaps — platforms with no instruction files at all.
- */
-function detectCoverageGaps(model) {
-  const drifts = [];
-  const allPlatforms = model.activePlatforms;
-
-  for (const ap of allPlatforms) {
-    const detail = model.platformDetails[ap.platform];
-    if (!detail) continue;
-
-    if (detail.instructionFiles.length === 0 || !detail.instructionContent.trim()) {
-      drifts.push({
-        type: 'coverage-gap',
-        severity: 'high',
-        platforms: [ap.platform],
-        description: `${ap.label} is detected but has no instruction content`,
-        recommendation: `Create instruction file for ${ap.label} to ensure consistent agent behavior.`,
-      });
-    }
-
-    if (detail.configFiles.length === 0) {
-      drifts.push({
-        type: 'coverage-gap',
-        severity: 'medium',
-        platforms: [ap.platform],
-        description: `${ap.label} has no configuration file`,
-        recommendation: `Add configuration for ${ap.label} to explicitly set trust and behavior parameters.`,
-      });
-    }
-  }
-
-  return drifts;
-}
-
-/**
- * Detect instruction drift — key instruction patterns present in some but not all platforms.
- */
-function detectInstructionDrift(model) {
-  const drifts = [];
-  const platforms = model.activePlatforms.map(p => p.platform);
-
-  if (platforms.length < 2) return drifts;
-
-  // Check for common instruction patterns that should be shared
-  const patterns = [
-    { name: 'test command', regex: /(?:test|pytest|jest|vitest|npm\s+test)/i, severity: 'high' },
-    { name: 'lint command', regex: /(?:lint|eslint|ruff|pylint)/i, severity: 'medium' },
-    { name: 'build command', regex: /(?:build|compile|tsc|webpack|vite build)/i, severity: 'medium' },
-    { name: 'architecture diagram', regex: /(?:mermaid|graph\s+TD|flowchart)/i, severity: 'low' },
-    { name: 'security instructions', regex: /(?:\.env|secrets?|credential|NEVER.*push|deny.*read)/i, severity: 'high' },
-    { name: 'language preference', regex: /(?:hebrew|english|language|speak|respond in)/i, severity: 'low' },
-  ];
-
-  for (const pattern of patterns) {
-    const has = [];
-    const hasNot = [];
-
-    for (const p of platforms) {
-      const detail = model.platformDetails[p];
-      if (!detail) continue;
-      const content = detail.instructionContent || '';
-      if (pattern.regex.test(content)) {
-        has.push(p);
-      } else {
-        hasNot.push(p);
-      }
-    }
-
-    if (has.length > 0 && hasNot.length > 0) {
-      drifts.push({
-        type: 'instruction-drift',
-        severity: pattern.severity,
-        platforms: [...has, ...hasNot],
-        description: `"${pattern.name}" found in ${has.join(', ')} but missing from ${hasNot.join(', ')}`,
-        recommendation: `Add ${pattern.name} instructions to ${hasNot.join(', ')} for consistent behavior.`,
-      });
-    }
-  }
-
-  return drifts;
-}
-
-// ─── Main detection function ────────────────────────────────────────────────
-
-/**
- * Detect all types of drift across platforms in a canonical model.
- *
- * @param {object} canonicalModel - Output of buildCanonicalModel()
- * @returns {object} { drifts, summary, harmonyScore }
- */
-function detectDrift(canonicalModel) {
-  const drifts = [
-    ...detectInstructionDrift(canonicalModel),
-    ...detectTrustDrift(canonicalModel),
-    ...detectMcpDrift(canonicalModel),
-    ...detectRuleDrift(canonicalModel),
-    ...detectCoverageGaps(canonicalModel),
-  ];
-
-  // Sort by severity (critical first)
-  drifts.sort((a, b) => (SEVERITY_ORDER[b.severity] || 0) - (SEVERITY_ORDER[a.severity] || 0));
-
-  // Summary counts
-  const summary = {
-    total: drifts.length,
-    critical: drifts.filter(d => d.severity === 'critical').length,
-    high: drifts.filter(d => d.severity === 'high').length,
-    medium: drifts.filter(d => d.severity === 'medium').length,
-    low: drifts.filter(d => d.severity === 'low').length,
-  };
-
-  // Harmony score: start at 100, deduct per drift
-  const deductions = {
-    critical: 20,
-    high: 12,
-    medium: 5,
-    low: 2,
-  };
-
-  let harmonyScore = 100;
-  for (const drift of drifts) {
-    harmonyScore -= deductions[drift.severity] || 2;
-  }
-  harmonyScore = Math.max(0, Math.min(100, harmonyScore));
-
-  return {
-    drifts,
-    summary,
-    harmonyScore,
-  };
-}
-
-// ─── Report formatter ───────────────────────────────────────────────────────
-
-const SEVERITY_ICONS = {
-  critical: '\u2718',  // ✘
-  high: '!',
-  medium: '~',
-  low: '-',
-};
-
-const SEVERITY_COLORS = {
-  critical: 'red',
-  high: 'yellow',
-  medium: 'blue',
-  low: 'dim',
-};
-
-/**
- * Format a drift report for console output.
- *
- * @param {object} driftResult - Output of detectDrift()
- * @param {object} [options] - { color: true, verbose: false }
- * @returns {string} Formatted report
- */
-function formatDriftReport(driftResult, options = {}) {
-  const { color = true, verbose = false } = options;
-  const c = color ? colorize : (text) => text;
-  const lines = [];
-
-  lines.push('');
-  lines.push(c('  Harmony Drift Report', 'bold'));
-  lines.push(c('  ' + '='.repeat(40), 'dim'));
-  lines.push('');
-
-  // Score
-  const score = driftResult.harmonyScore;
-  const scoreColor = score >= 80 ? 'green' : score >= 50 ? 'yellow' : 'red';
-  lines.push(`  Harmony Score: ${c(String(score) + '/100', scoreColor)}`);
-  lines.push('');
-
-  // Summary
-  const s = driftResult.summary;
-  if (s.total === 0) {
-    lines.push(c('  No drift detected. All platforms are aligned.', 'green'));
-  } else {
-    lines.push(`  Drift items: ${s.total} total`);
-    if (s.critical > 0) lines.push(c(`    Critical: ${s.critical}`, 'red'));
-    if (s.high > 0) lines.push(c(`    High:     ${s.high}`, 'yellow'));
-    if (s.medium > 0) lines.push(c(`    Medium:   ${s.medium}`, 'blue'));
-    if (s.low > 0) lines.push(c(`    Low:      ${s.low}`, 'dim'));
-  }
-
-  lines.push('');
-
-  // Individual drifts
-  for (const drift of driftResult.drifts) {
-    const icon = SEVERITY_ICONS[drift.severity] || '-';
-    const dColor = SEVERITY_COLORS[drift.severity] || 'dim';
-
-    lines.push(c(`  ${icon} [${drift.severity.toUpperCase()}] ${drift.type}`, dColor));
-    lines.push(`    ${drift.description}`);
-    if (verbose && drift.recommendation) {
-      lines.push(c(`    Recommendation: ${drift.recommendation}`, 'dim'));
-    }
-    lines.push('');
-  }
-
-  return lines.join('\n');
-}
-
-module.exports = {
-  detectDrift,
-  formatDriftReport,
-};
+/**
+ * Harmony Drift Detection Engine
+ *
+ * Compares configurations ACROSS platforms and finds inconsistencies
+ * in instructions, trust posture, MCP servers, rules, and coverage.
+ *
+ * Drift types:
+ *   instruction-drift  — different platforms say different things
+ *   trust-drift        — sandbox/approval modes differ across platforms
+ *   mcp-drift          — MCP servers not aligned across platforms
+ *   rule-drift         — rule coverage differs between platforms
+ *   coverage-gap       — platform missing instructions entirely
+ *
+ * Severity: critical, high, medium, low
+ */
+
+const SEVERITY_ORDER = { critical: 3, high: 2, medium: 1, low: 0 };
+
+const COLORS = {
+  reset: '\x1b[0m',
+  bold: '\x1b[1m',
+  dim: '\x1b[2m',
+  red: '\x1b[31m',
+  green: '\x1b[32m',
+  yellow: '\x1b[33m',
+  blue: '\x1b[36m',
+  magenta: '\x1b[35m',
+};
+
+function colorize(text, color) {
+  return `${COLORS[color] || ''}${text}${COLORS.reset}`;
+}
+
+// ─── Trust level risk mapping ───────────────────────────────────────────────
+
+const TRUST_RISK = {
+  'bypass': 4,
+  'full-auto': 4,
+  'unrestricted': 4,
+  'no-sandbox': 3,
+  'standard': 2,
+  'safe-write': 2,
+  'default': 1,
+  'locked-down': 0,
+  'unknown': 1,
+};
+
+// ─── Drift detectors ───────────────────────────────────────────────────────
+
+/**
+ * Detect trust posture drift across platforms.
+ * Critical if risk gap >= 2 levels between any two platforms.
+ */
+function detectTrustDrift(model) {
+  const drifts = [];
+  const platforms = Object.keys(model.trustPosture);
+
+  if (platforms.length < 2) return drifts;
+
+  const riskLevels = {};
+  for (const p of platforms) {
+    riskLevels[p] = TRUST_RISK[model.trustPosture[p]] ?? 1;
+  }
+
+  const maxRisk = Math.max(...Object.values(riskLevels));
+  const minRisk = Math.min(...Object.values(riskLevels));
+  const gap = maxRisk - minRisk;
+
+  if (gap === 0) return drifts;
+
+  const highRiskPlatforms = platforms.filter(p => riskLevels[p] === maxRisk);
+  const lowRiskPlatforms = platforms.filter(p => riskLevels[p] === minRisk);
+
+  let severity = 'low';
+  if (gap >= 3) severity = 'critical';
+  else if (gap >= 2) severity = 'high';
+  else if (gap >= 1) severity = 'medium';
+
+  drifts.push({
+    type: 'trust-drift',
+    severity,
+    platforms,
+    description: `Trust posture gap of ${gap} levels: ` +
+      `${highRiskPlatforms.map(p => `${p}=${model.trustPosture[p]}`).join(', ')} vs ` +
+      `${lowRiskPlatforms.map(p => `${p}=${model.trustPosture[p]}`).join(', ')}`,
+    recommendation: `Align trust posture across platforms. Consider raising ${lowRiskPlatforms.join(', ')} ` +
+      `or tightening ${highRiskPlatforms.join(', ')} to reduce risk surface.`,
+  });
+
+  return drifts;
+}
+
+/**
+ * Detect MCP server drift — servers present on some platforms but not others.
+ */
+function detectMcpDrift(model) {
+  const drifts = [];
+  const allPlatforms = model.activePlatforms.map(p => p.platform);
+  const mcpCapablePlatforms = allPlatforms.filter(p =>
+    p === 'claude' || p === 'gemini' || p === 'copilot' || p === 'cursor' || p === 'windsurf' || p === 'opencode'
+  );
+
+  if (mcpCapablePlatforms.length < 2) return drifts;
+
+  for (const [name, server] of Object.entries(model.mcpServers)) {
+    const present = server.platforms.filter(p => mcpCapablePlatforms.includes(p));
+    const missing = mcpCapablePlatforms.filter(p => !server.platforms.includes(p));
+
+    if (missing.length > 0 && present.length > 0) {
+      drifts.push({
+        type: 'mcp-drift',
+        severity: missing.length >= 2 ? 'high' : 'medium',
+        platforms: [...present, ...missing],
+        description: `MCP server "${name}" is configured on ${present.join(', ')} but missing from ${missing.join(', ')}`,
+        recommendation: `Add "${name}" to ${missing.join(', ')} MCP configuration to ensure consistent tool access.`,
+      });
+    }
+  }
+
+  // Check for platforms with zero MCP servers when others have some
+  const serverCounts = {};
+  for (const p of mcpCapablePlatforms) {
+    const detail = model.platformDetails[p];
+    serverCounts[p] = detail ? detail.mcpServers.length : 0;
+  }
+
+  const withServers = mcpCapablePlatforms.filter(p => serverCounts[p] > 0);
+  const withoutServers = mcpCapablePlatforms.filter(p => serverCounts[p] === 0);
+
+  if (withServers.length > 0 && withoutServers.length > 0) {
+    drifts.push({
+      type: 'mcp-drift',
+      severity: 'medium',
+      platforms: [...withServers, ...withoutServers],
+      description: `MCP servers configured on ${withServers.join(', ')} (${withServers.map(p => serverCounts[p]).join(', ')} servers) ` +
+        `but ${withoutServers.join(', ')} have none`,
+      recommendation: `Consider adding MCP servers to ${withoutServers.join(', ')} for consistent tool access.`,
+    });
+  }
+
+  return drifts;
+}
+
+/**
+ * Detect rule drift — platforms with rules configured vs those without.
+ */
+function detectRuleDrift(model) {
+  const drifts = [];
+  const allPlatforms = model.activePlatforms.map(p => p.platform);
+
+  // Platforms that support rules: claude (.claude/rules), copilot (.github/instructions), cursor (.cursor/rules)
+  const ruleCapable = allPlatforms.filter(p =>
+    p === 'claude' || p === 'copilot' || p === 'cursor' || p === 'windsurf'
+  );
+
+  if (ruleCapable.length < 2) return drifts;
+
+  const ruleCounts = {};
+  for (const p of ruleCapable) {
+    ruleCounts[p] = model.activePlatforms.find(ap => ap.platform === p)?.ruleCount || 0;
+  }
+
+  const withRules = ruleCapable.filter(p => ruleCounts[p] > 0);
+  const withoutRules = ruleCapable.filter(p => ruleCounts[p] === 0);
+
+  if (withRules.length > 0 && withoutRules.length > 0) {
+    drifts.push({
+      type: 'rule-drift',
+      severity: 'medium',
+      platforms: [...withRules, ...withoutRules],
+      description: `Rules configured on ${withRules.map(p => `${p} (${ruleCounts[p]})`).join(', ')} ` +
+        `but missing from ${withoutRules.join(', ')}`,
+      recommendation: `Add equivalent rules to ${withoutRules.join(', ')} to maintain consistent behavior constraints.`,
+    });
+  }
+
+  // Check for large rule count gaps
+  const maxRules = Math.max(...Object.values(ruleCounts));
+  const minRules = Math.min(...Object.values(ruleCounts));
+
+  if (maxRules > 0 && minRules > 0 && maxRules > minRules * 3) {
+    const richPlatform = ruleCapable.find(p => ruleCounts[p] === maxRules);
+    const sparPlatform = ruleCapable.find(p => ruleCounts[p] === minRules);
+    drifts.push({
+      type: 'rule-drift',
+      severity: 'low',
+      platforms: [richPlatform, sparPlatform],
+      description: `Large rule count gap: ${richPlatform} has ${maxRules} rules vs ${sparPlatform} with ${minRules}`,
+      recommendation: `Review whether ${sparPlatform} is missing important rules that ${richPlatform} enforces.`,
+    });
+  }
+
+  return drifts;
+}
+
+/**
+ * Detect instruction coverage gaps — platforms with no instruction files at all.
+ */
+function detectCoverageGaps(model) {
+  const drifts = [];
+  const allPlatforms = model.activePlatforms;
+
+  for (const ap of allPlatforms) {
+    const detail = model.platformDetails[ap.platform];
+    if (!detail) continue;
+
+    if (detail.instructionFiles.length === 0 || !detail.instructionContent.trim()) {
+      drifts.push({
+        type: 'coverage-gap',
+        severity: 'high',
+        platforms: [ap.platform],
+        description: `${ap.label} is detected but has no instruction content`,
+        recommendation: `Create instruction file for ${ap.label} to ensure consistent agent behavior.`,
+      });
+    }
+
+    if (detail.configFiles.length === 0) {
+      drifts.push({
+        type: 'coverage-gap',
+        severity: 'medium',
+        platforms: [ap.platform],
+        description: `${ap.label} has no configuration file`,
+        recommendation: `Add configuration for ${ap.label} to explicitly set trust and behavior parameters.`,
+      });
+    }
+  }
+
+  return drifts;
+}
+
+/**
+ * Detect instruction drift — key instruction patterns present in some but not all platforms.
+ */
+function detectInstructionDrift(model) {
+  const drifts = [];
+  const platforms = model.activePlatforms.map(p => p.platform);
+
+  if (platforms.length < 2) return drifts;
+
+  // Check for common instruction patterns that should be shared
+  const patterns = [
+    { name: 'test command', regex: /(?:test|pytest|jest|vitest|npm\s+test)/i, severity: 'high' },
+    { name: 'lint command', regex: /(?:lint|eslint|ruff|pylint)/i, severity: 'medium' },
+    { name: 'build command', regex: /(?:build|compile|tsc|webpack|vite build)/i, severity: 'medium' },
+    { name: 'architecture diagram', regex: /(?:mermaid|graph\s+TD|flowchart)/i, severity: 'low' },
+    { name: 'security instructions', regex: /(?:\.env|secrets?|credential|NEVER.*push|deny.*read)/i, severity: 'high' },
+    { name: 'language preference', regex: /(?:hebrew|english|language|speak|respond in)/i, severity: 'low' },
+  ];
+
+  for (const pattern of patterns) {
+    const has = [];
+    const hasNot = [];
+
+    for (const p of platforms) {
+      const detail = model.platformDetails[p];
+      if (!detail) continue;
+      const content = detail.instructionContent || '';
+      if (pattern.regex.test(content)) {
+        has.push(p);
+      } else {
+        hasNot.push(p);
+      }
+    }
+
+    if (has.length > 0 && hasNot.length > 0) {
+      drifts.push({
+        type: 'instruction-drift',
+        severity: pattern.severity,
+        platforms: [...has, ...hasNot],
+        description: `"${pattern.name}" found in ${has.join(', ')} but missing from ${hasNot.join(', ')}`,
+        recommendation: `Add ${pattern.name} instructions to ${hasNot.join(', ')} for consistent behavior.`,
+      });
+    }
+  }
+
+  return drifts;
+}
+
+// ─── Main detection function ────────────────────────────────────────────────
+
+/**
+ * Detect all types of drift across platforms in a canonical model.
+ *
+ * @param {object} canonicalModel - Output of buildCanonicalModel()
+ * @returns {object} { drifts, summary, harmonyScore }
+ */
+function detectDrift(canonicalModel) {
+  const drifts = [
+    ...detectInstructionDrift(canonicalModel),
+    ...detectTrustDrift(canonicalModel),
+    ...detectMcpDrift(canonicalModel),
+    ...detectRuleDrift(canonicalModel),
+    ...detectCoverageGaps(canonicalModel),
+  ];
+
+  // Sort by severity (critical first)
+  drifts.sort((a, b) => (SEVERITY_ORDER[b.severity] || 0) - (SEVERITY_ORDER[a.severity] || 0));
+
+  // Summary counts
+  const summary = {
+    total: drifts.length,
+    critical: drifts.filter(d => d.severity === 'critical').length,
+    high: drifts.filter(d => d.severity === 'high').length,
+    medium: drifts.filter(d => d.severity === 'medium').length,
+    low: drifts.filter(d => d.severity === 'low').length,
+  };
+
+  // Harmony score: start at 100, deduct per drift
+  const deductions = {
+    critical: 20,
+    high: 12,
+    medium: 5,
+    low: 2,
+  };
+
+  let harmonyScore = 100;
+  for (const drift of drifts) {
+    harmonyScore -= deductions[drift.severity] || 2;
+  }
+  harmonyScore = Math.max(0, Math.min(100, harmonyScore));
+
+  return {
+    drifts,
+    summary,
+    harmonyScore,
+  };
+}
+
+// ─── Report formatter ───────────────────────────────────────────────────────
+
+const SEVERITY_ICONS = {
+  critical: '\u2718',  // ✘
+  high: '!',
+  medium: '~',
+  low: '-',
+};
+
+const SEVERITY_COLORS = {
+  critical: 'red',
+  high: 'yellow',
+  medium: 'blue',
+  low: 'dim',
+};
+
+/**
+ * Format a drift report for console output.
+ *
+ * @param {object} driftResult - Output of detectDrift()
+ * @param {object} [options] - { color: true, verbose: false }
+ * @returns {string} Formatted report
+ */
+function formatDriftReport(driftResult, options = {}) {
+  const { color = true, verbose = false } = options;
+  const c = color ? colorize : (text) => text;
+  const lines = [];
+
+  lines.push('');
+  lines.push(c('  Harmony Drift Report', 'bold'));
+  lines.push(c('  ' + '='.repeat(40), 'dim'));
+  lines.push('');
+
+  // Score
+  const score = driftResult.harmonyScore;
+  const scoreColor = score >= 80 ? 'green' : score >= 50 ? 'yellow' : 'red';
+  lines.push(`  Harmony Score: ${c(String(score) + '/100', scoreColor)}`);
+  lines.push('');
+
+  // Summary
+  const s = driftResult.summary;
+  if (s.total === 0) {
+    lines.push(c('  No drift detected. All platforms are aligned.', 'green'));
+  } else {
+    lines.push(`  Drift items: ${s.total} total`);
+    if (s.critical > 0) lines.push(c(`    Critical: ${s.critical}`, 'red'));
+    if (s.high > 0) lines.push(c(`    High:     ${s.high}`, 'yellow'));
+    if (s.medium > 0) lines.push(c(`    Medium:   ${s.medium}`, 'blue'));
+    if (s.low > 0) lines.push(c(`    Low:      ${s.low}`, 'dim'));
+  }
+
+  lines.push('');
+
+  // Individual drifts
+  for (const drift of driftResult.drifts) {
+    const icon = SEVERITY_ICONS[drift.severity] || '-';
+    const dColor = SEVERITY_COLORS[drift.severity] || 'dim';
+
+    lines.push(c(`  ${icon} [${drift.severity.toUpperCase()}] ${drift.type}`, dColor));
+    lines.push(`    ${drift.description}`);
+    if (verbose && drift.recommendation) {
+      lines.push(c(`    Recommendation: ${drift.recommendation}`, 'dim'));
+    }
+    lines.push('');
+  }
+
+  return lines.join('\n');
+}
+
+module.exports = {
+  detectDrift,
+  formatDriftReport,
+};