@nerviq/cli 1.29.0 → 1.30.0

package/src/gemini/activity.js

@@ -1,402 +1,402 @@
-/**
- * Gemini Repeat-Usage Surfaces
- *
- * Adapts the shared activity/snapshot backend for Gemini CLI platform.
- * Provides: history, compare, trend, feedback, insights.
- *
- * Gemini snapshots are stored alongside Claude snapshots in
- * .nerviq/snapshots/ (legacy: .claude/nerviq-cli/snapshots/) but filtered by platform='gemini'.
- */
-
-const path = require('path');
-const {
-  readSnapshotIndex,
-  recordRecommendationOutcome,
-  readOutcomeIndex,
-  summarizeOutcomeEntries,
-} = require('../activity');
-const { version } = require('../../package.json');
-
-// --- History ---
-
-/**
- * Get Gemini audit history from snapshots.
- * Filters to platform='gemini' snapshots only.
- */
-function getGeminiHistory(dir, limit = 20) {
-  const entries = readSnapshotIndex(dir);
-  return entries
-    .filter(e => e.snapshotKind === 'audit' && (e.platform === 'gemini' || e.summary?.platform === 'gemini'))
-    .sort((a, b) => new Date(b.createdAt) - new Date(a.createdAt))
-    .slice(0, limit);
-}
-
-function formatGeminiHistory(dir) {
-  const history = getGeminiHistory(dir, 10);
-  if (history.length === 0) {
-    return 'No Gemini snapshots found. Run `npx nerviq --platform gemini --snapshot` to save one.';
-  }
-
-  const lines = ['Gemini Score History (most recent first):', ''];
-  for (const entry of history) {
-    const date = entry.createdAt?.split('T')[0] || 'unknown';
-    const score = entry.summary?.score ?? '?';
-    const passed = entry.summary?.passed ?? '?';
-    const total = entry.summary?.checkCount ?? '?';
-    lines.push(`  ${date}  ${score}/100  (${passed}/${total} passing)`);
-  }
-
-  const comparison = compareGeminiLatest(dir);
-  if (comparison) {
-    lines.push('');
-    const sign = comparison.delta.score >= 0 ? '+' : '';
-    lines.push(`  Trend: ${comparison.trend} (${sign}${comparison.delta.score} since previous)`);
-    if (comparison.improvements.length > 0) {
-      lines.push(`  Fixed: ${comparison.improvements.join(', ')}`);
-    }
-    if (comparison.regressions.length > 0) {
-      lines.push(`  New gaps: ${comparison.regressions.join(', ')}`);
-    }
-  }
-
-  return lines.join('\n');
-}
-
-// --- Compare ---
-
-/**
- * Compare the two most recent Gemini audit snapshots.
- */
-function compareGeminiLatest(dir) {
-  const audits = getGeminiHistory(dir, 2);
-  if (audits.length < 2) return null;
-
-  const current = audits[0];
-  const previous = audits[1];
-
-  const delta = {
-    score: (current.summary?.score || 0) - (previous.summary?.score || 0),
-    organic: (current.summary?.organicScore || 0) - (previous.summary?.organicScore || 0),
-    passed: (current.summary?.passed || 0) - (previous.summary?.passed || 0),
-  };
-
-  const regressions = [];
-  const improvements = [];
-
-  const prevKeys = new Set(previous.summary?.topActionKeys || []);
-  const currKeys = new Set(current.summary?.topActionKeys || []);
-
-  for (const key of currKeys) {
-    if (!prevKeys.has(key)) regressions.push(key);
-  }
-  for (const key of prevKeys) {
-    if (!currKeys.has(key)) improvements.push(key);
-  }
-
-  return {
-    platform: 'gemini',
-    current: { date: current.createdAt, score: current.summary?.score, passed: current.summary?.passed },
-    previous: { date: previous.createdAt, score: previous.summary?.score, passed: previous.summary?.passed },
-    delta,
-    regressions,
-    improvements,
-    trend: delta.score > 0 ? 'improving' : delta.score < 0 ? 'regressing' : 'stable',
-  };
-}
-
-// --- Trend ---
-
-/**
- * Export Gemini trend report as markdown.
- */
-function exportGeminiTrendReport(dir) {
-  const history = getGeminiHistory(dir, 50);
-  if (history.length === 0) return null;
-
-  const comparison = compareGeminiLatest(dir);
-  const lines = [
-    '# Gemini CLI Setup Trend Report',
-    '',
-    `**Project:** ${path.basename(dir)}`,
-    `**Platform:** Gemini CLI`,
-    `**Generated:** ${new Date().toISOString().split('T')[0]}`,
-    `**Snapshots:** ${history.length}`,
-    '',
-    '## Score History',
-    '',
-    '| Date | Score | Passed | Checks |',
-    '|------|-------|--------|--------|',
-  ];
-
-  for (const entry of history) {
-    const date = entry.createdAt?.split('T')[0] || '?';
-    lines.push(`| ${date} | ${entry.summary?.score ?? '?'}/100 | ${entry.summary?.passed ?? '?'} | ${entry.summary?.checkCount ?? '?'} |`);
-  }
-
-  if (comparison) {
-    lines.push('');
-    lines.push('## Latest Comparison');
-    lines.push('');
-    lines.push(`- **Previous:** ${comparison.previous.score}/100 (${comparison.previous.date?.split('T')[0]})`);
-    lines.push(`- **Current:** ${comparison.current.score}/100 (${comparison.current.date?.split('T')[0]})`);
-    lines.push(`- **Delta:** ${comparison.delta.score >= 0 ? '+' : ''}${comparison.delta.score} points`);
-    lines.push(`- **Trend:** ${comparison.trend}`);
-    if (comparison.improvements.length > 0) lines.push(`- **Fixed:** ${comparison.improvements.join(', ')}`);
-    if (comparison.regressions.length > 0) lines.push(`- **New gaps:** ${comparison.regressions.join(', ')}`);
-  }
-
-  // ASCII trend chart
-  if (history.length >= 3) {
-    lines.push('');
-    lines.push('## Trend Chart');
-    lines.push('');
-    lines.push('```');
-    const scores = history.slice().reverse().map(e => e.summary?.score ?? 0);
-    const max = Math.max(...scores, 100);
-    const chartHeight = 10;
-    for (let row = chartHeight; row >= 0; row--) {
-      const threshold = (row / chartHeight) * max;
-      const rowLabel = String(Math.round(threshold)).padStart(3);
-      const bar = scores.map(s => s >= threshold ? '#' : ' ').join('');
-      lines.push(`${rowLabel} |${bar}`);
-    }
-    lines.push(`    +${'─'.repeat(scores.length)}`);
-    lines.push('```');
-  }
-
-  lines.push('');
-  lines.push(`---`);
-  lines.push(`*Generated by nerviq v${version} for Gemini CLI*`);
-  return lines.join('\n');
-}
-
-// --- Feedback ---
-
-/**
- * Record feedback on a Gemini recommendation.
- */
-function recordGeminiFeedback(dir, payload) {
-  return recordRecommendationOutcome(dir, {
-    ...payload,
-    source: payload.source || 'gemini-cli',
-    platform: 'gemini',
-  });
-}
-
-/**
- * Get Gemini feedback summary.
- */
-function getGeminiFeedbackSummary(dir) {
-  const entries = readOutcomeIndex(dir)
-    .filter(e => e.source === 'gemini-cli' || e.platform === 'gemini');
-  return summarizeOutcomeEntries(entries);
-}
-
-function formatGeminiFeedback(dir) {
-  const summary = getGeminiFeedbackSummary(dir);
-  if (!summary || Object.keys(summary).length === 0) {
-    return 'No Gemini feedback recorded yet. Use `npx nerviq --platform gemini feedback` to rate recommendations.';
-  }
-
-  const lines = ['Gemini Recommendation Feedback:', ''];
-  const entries = Array.isArray(summary) ? summary : Object.values(summary);
-  for (const entry of entries) {
-    const key = entry.key || 'unknown';
-    const accepted = entry.accepted || 0;
-    const rejected = entry.rejected || 0;
-    const total = entry.total || 0;
-    lines.push(`  ${key}: ${accepted} accepted, ${rejected} rejected (${total} total)`);
-  }
-  return lines.join('\n');
-}
-
-// --- Insights ---
-
-/**
- * Generate Gemini-specific insights from audit history and feedback.
- * Includes Gemini-unique patterns: sandbox-drift, policy-complexity,
- * caveat-persistence, trust-class-stagnation.
- */
-function generateGeminiInsights(dir) {
-  const history = getGeminiHistory(dir, 50);
-  const feedback = getGeminiFeedbackSummary(dir);
-  const insights = [];
-
-  // Pattern 1: Persistent failures
-  if (history.length >= 3) {
-    const recentFailKeys = new Map();
-    for (const entry of history.slice(0, 5)) {
-      for (const key of (entry.summary?.topActionKeys || [])) {
-        recentFailKeys.set(key, (recentFailKeys.get(key) || 0) + 1);
-      }
-    }
-    for (const [key, count] of recentFailKeys) {
-      if (count >= 3) {
-        insights.push({
-          type: 'persistent-failure',
-          severity: 'high',
-          key,
-          message: `Check ${key} has failed in ${count} of the last ${Math.min(history.length, 5)} audits. Consider addressing it or marking it as intentionally skipped.`,
-        });
-      }
-    }
-  }
-
-  // Pattern 2: Score regression
-  if (history.length >= 2) {
-    const scores = history.map(e => e.summary?.score ?? 0);
-    if (scores[0] < scores[1]) {
-      insights.push({
-        type: 'regression-pattern',
-        severity: 'medium',
-        message: `Score dropped from ${scores[1]} to ${scores[0]} in the most recent audit. Review recent changes.`,
-        delta: scores[0] - scores[1],
-      });
-    }
-  }
-
-  // Pattern 3: Improvement velocity stall
-  if (history.length >= 5) {
-    const recentScores = history.slice(0, 5).map(e => e.summary?.score ?? 0);
-    const range = Math.max(...recentScores) - Math.min(...recentScores);
-    if (range <= 2) {
-      insights.push({
-        type: 'velocity-stall',
-        severity: 'low',
-        message: `Score has been flat (range: ${range} points) over the last 5 audits. Consider addressing lower-priority checks.`,
-      });
-    }
-  }
-
-  // Pattern 4: Feedback signals
-  const feedbackEntries = Array.isArray(feedback) ? feedback : Object.values(feedback || {});
-  for (const entry of feedbackEntries) {
-    if (entry.rejected > entry.accepted && entry.total >= 2) {
-      insights.push({
-        type: 'feedback-signal',
-        severity: 'medium',
-        key: entry.key,
-        message: `Recommendation ${entry.key} has been rejected more than accepted (${entry.rejected}/${entry.total}). Consider adjusting or removing this recommendation.`,
-      });
-    }
-  }
-
-  // Gemini-specific Pattern 5: Sandbox drift
-  if (history.length >= 3) {
-    const sandboxKeys = [];
-    for (const entry of history.slice(0, 5)) {
-      for (const key of (entry.summary?.topActionKeys || [])) {
-        if (key.includes('sandbox') || key.includes('isolation')) {
-          sandboxKeys.push(key);
-        }
-      }
-    }
-    if (sandboxKeys.length >= 2) {
-      insights.push({
-        type: 'sandbox-drift',
-        severity: 'high',
-        message: `Sandbox-related checks have appeared in ${sandboxKeys.length} recent audits. Gemini sandbox configuration may be drifting from recommended posture.`,
-        keys: [...new Set(sandboxKeys)],
-      });
-    }
-  }
-
-  // Gemini-specific Pattern 6: Policy complexity
-  if (history.length >= 2) {
-    const latest = history[0];
-    const policyKeys = (latest.summary?.topActionKeys || []).filter(
-      k => k.includes('policy') || k.includes('governance') || k.includes('rule')
-    );
-    if (policyKeys.length >= 3) {
-      insights.push({
-        type: 'policy-complexity',
-        severity: 'medium',
-        message: `${policyKeys.length} policy/governance checks are failing simultaneously. Gemini policy configuration may be overly complex or misconfigured.`,
-        keys: policyKeys,
-      });
-    }
-  }
-
-  // Gemini-specific Pattern 7: Caveat persistence
-  if (history.length >= 3) {
-    const caveatKeys = new Map();
-    for (const entry of history.slice(0, 5)) {
-      for (const key of (entry.summary?.topActionKeys || [])) {
-        if (key.includes('caveat') || key.includes('warning') || key.includes('limitation')) {
-          caveatKeys.set(key, (caveatKeys.get(key) || 0) + 1);
-        }
-      }
-    }
-    for (const [key, count] of caveatKeys) {
-      if (count >= 3) {
-        insights.push({
-          type: 'caveat-persistence',
-          severity: 'medium',
-          key,
-          message: `Caveat/limitation ${key} has persisted across ${count} audits. This may indicate a Gemini CLI limitation that should be documented or worked around.`,
-        });
-      }
-    }
-  }
-
-  // Gemini-specific Pattern 8: Trust class stagnation
-  if (history.length >= 5) {
-    const trustKeys = new Map();
-    for (const entry of history.slice(0, 5)) {
-      for (const key of (entry.summary?.topActionKeys || [])) {
-        if (key.includes('trust') || key.includes('approval') || key.includes('permission')) {
-          trustKeys.set(key, (trustKeys.get(key) || 0) + 1);
-        }
-      }
-    }
-    for (const [key, count] of trustKeys) {
-      if (count >= 4) {
-        insights.push({
-          type: 'trust-class-stagnation',
-          severity: 'medium',
-          key,
-          message: `Trust/permission check ${key} has been failing in ${count} of the last 5 audits. Gemini trust classification may need elevation or explicit override.`,
-        });
-      }
-    }
-  }
-
-  return {
-    platform: 'gemini',
-    generatedAt: new Date().toISOString(),
-    snapshotCount: history.length,
-    feedbackCount: feedbackEntries.length,
-    insights,
-    summary: insights.length === 0
-      ? 'No actionable insights detected. Keep running audits to build pattern data.'
-      : `${insights.length} insight(s) detected across ${history.length} snapshots.`,
-  };
-}
-
-function formatGeminiInsights(dir) {
-  const result = generateGeminiInsights(dir);
-  if (result.insights.length === 0) {
-    return result.summary;
-  }
-
-  const lines = ['Gemini Insights:', ''];
-  for (const insight of result.insights) {
-    const severity = insight.severity.toUpperCase();
-    lines.push(`  [${severity}] ${insight.message}`);
-  }
-  lines.push('');
-  lines.push(result.summary);
-  return lines.join('\n');
-}
-
-module.exports = {
-  getGeminiHistory,
-  formatGeminiHistory,
-  compareGeminiLatest,
-  exportGeminiTrendReport,
-  recordGeminiFeedback,
-  getGeminiFeedbackSummary,
-  formatGeminiFeedback,
-  generateGeminiInsights,
-  formatGeminiInsights,
-};
+/**
+ * Gemini Repeat-Usage Surfaces
+ *
+ * Adapts the shared activity/snapshot backend for Gemini CLI platform.
+ * Provides: history, compare, trend, feedback, insights.
+ *
+ * Gemini snapshots are stored alongside Claude snapshots in
+ * .nerviq/snapshots/ (legacy: .claude/nerviq-cli/snapshots/) but filtered by platform='gemini'.
+ */
+
+const path = require('path');
+const {
+  readSnapshotIndex,
+  recordRecommendationOutcome,
+  readOutcomeIndex,
+  summarizeOutcomeEntries,
+} = require('../activity');
+const { version } = require('../../package.json');
+
+// --- History ---
+
+/**
+ * Get Gemini audit history from snapshots.
+ * Filters to platform='gemini' snapshots only.
+ */
+function getGeminiHistory(dir, limit = 20) {
+  const entries = readSnapshotIndex(dir);
+  return entries
+    .filter(e => e.snapshotKind === 'audit' && (e.platform === 'gemini' || e.summary?.platform === 'gemini'))
+    .sort((a, b) => new Date(b.createdAt) - new Date(a.createdAt))
+    .slice(0, limit);
+}
+
+function formatGeminiHistory(dir) {
+  const history = getGeminiHistory(dir, 10);
+  if (history.length === 0) {
+    return 'No Gemini snapshots found. Run `npx nerviq --platform gemini --snapshot` to save one.';
+  }
+
+  const lines = ['Gemini Score History (most recent first):', ''];
+  for (const entry of history) {
+    const date = entry.createdAt?.split('T')[0] || 'unknown';
+    const score = entry.summary?.score ?? '?';
+    const passed = entry.summary?.passed ?? '?';
+    const total = entry.summary?.checkCount ?? '?';
+    lines.push(`  ${date}  ${score}/100  (${passed}/${total} passing)`);
+  }
+
+  const comparison = compareGeminiLatest(dir);
+  if (comparison) {
+    lines.push('');
+    const sign = comparison.delta.score >= 0 ? '+' : '';
+    lines.push(`  Trend: ${comparison.trend} (${sign}${comparison.delta.score} since previous)`);
+    if (comparison.improvements.length > 0) {
+      lines.push(`  Fixed: ${comparison.improvements.join(', ')}`);
+    }
+    if (comparison.regressions.length > 0) {
+      lines.push(`  New gaps: ${comparison.regressions.join(', ')}`);
+    }
+  }
+
+  return lines.join('\n');
+}
+
+// --- Compare ---
+
+/**
+ * Compare the two most recent Gemini audit snapshots.
+ */
+function compareGeminiLatest(dir) {
+  const audits = getGeminiHistory(dir, 2);
+  if (audits.length < 2) return null;
+
+  const current = audits[0];
+  const previous = audits[1];
+
+  const delta = {
+    score: (current.summary?.score || 0) - (previous.summary?.score || 0),
+    organic: (current.summary?.organicScore || 0) - (previous.summary?.organicScore || 0),
+    passed: (current.summary?.passed || 0) - (previous.summary?.passed || 0),
+  };
+
+  const regressions = [];
+  const improvements = [];
+
+  const prevKeys = new Set(previous.summary?.topActionKeys || []);
+  const currKeys = new Set(current.summary?.topActionKeys || []);
+
+  for (const key of currKeys) {
+    if (!prevKeys.has(key)) regressions.push(key);
+  }
+  for (const key of prevKeys) {
+    if (!currKeys.has(key)) improvements.push(key);
+  }
+
+  return {
+    platform: 'gemini',
+    current: { date: current.createdAt, score: current.summary?.score, passed: current.summary?.passed },
+    previous: { date: previous.createdAt, score: previous.summary?.score, passed: previous.summary?.passed },
+    delta,
+    regressions,
+    improvements,
+    trend: delta.score > 0 ? 'improving' : delta.score < 0 ? 'regressing' : 'stable',
+  };
+}
+
+// --- Trend ---
+
+/**
+ * Export Gemini trend report as markdown.
+ */
+function exportGeminiTrendReport(dir) {
+  const history = getGeminiHistory(dir, 50);
+  if (history.length === 0) return null;
+
+  const comparison = compareGeminiLatest(dir);
+  const lines = [
+    '# Gemini CLI Setup Trend Report',
+    '',
+    `**Project:** ${path.basename(dir)}`,
+    `**Platform:** Gemini CLI`,
+    `**Generated:** ${new Date().toISOString().split('T')[0]}`,
+    `**Snapshots:** ${history.length}`,
+    '',
+    '## Score History',
+    '',
+    '| Date | Score | Passed | Checks |',
+    '|------|-------|--------|--------|',
+  ];
+
+  for (const entry of history) {
+    const date = entry.createdAt?.split('T')[0] || '?';
+    lines.push(`| ${date} | ${entry.summary?.score ?? '?'}/100 | ${entry.summary?.passed ?? '?'} | ${entry.summary?.checkCount ?? '?'} |`);
+  }
+
+  if (comparison) {
+    lines.push('');
+    lines.push('## Latest Comparison');
+    lines.push('');
+    lines.push(`- **Previous:** ${comparison.previous.score}/100 (${comparison.previous.date?.split('T')[0]})`);
+    lines.push(`- **Current:** ${comparison.current.score}/100 (${comparison.current.date?.split('T')[0]})`);
+    lines.push(`- **Delta:** ${comparison.delta.score >= 0 ? '+' : ''}${comparison.delta.score} points`);
+    lines.push(`- **Trend:** ${comparison.trend}`);
+    if (comparison.improvements.length > 0) lines.push(`- **Fixed:** ${comparison.improvements.join(', ')}`);
+    if (comparison.regressions.length > 0) lines.push(`- **New gaps:** ${comparison.regressions.join(', ')}`);
+  }
+
+  // ASCII trend chart
+  if (history.length >= 3) {
+    lines.push('');
+    lines.push('## Trend Chart');
+    lines.push('');
+    lines.push('```');
+    const scores = history.slice().reverse().map(e => e.summary?.score ?? 0);
+    const max = Math.max(...scores, 100);
+    const chartHeight = 10;
+    for (let row = chartHeight; row >= 0; row--) {
+      const threshold = (row / chartHeight) * max;
+      const rowLabel = String(Math.round(threshold)).padStart(3);
+      const bar = scores.map(s => s >= threshold ? '#' : ' ').join('');
+      lines.push(`${rowLabel} |${bar}`);
+    }
+    lines.push(`    +${'─'.repeat(scores.length)}`);
+    lines.push('```');
+  }
+
+  lines.push('');
+  lines.push(`---`);
+  lines.push(`*Generated by nerviq v${version} for Gemini CLI*`);
+  return lines.join('\n');
+}
+
+// --- Feedback ---
+
+/**
+ * Record feedback on a Gemini recommendation.
+ */
+function recordGeminiFeedback(dir, payload) {
+  return recordRecommendationOutcome(dir, {
+    ...payload,
+    source: payload.source || 'gemini-cli',
+    platform: 'gemini',
+  });
+}
+
+/**
+ * Get Gemini feedback summary.
+ */
+function getGeminiFeedbackSummary(dir) {
+  const entries = readOutcomeIndex(dir)
+    .filter(e => e.source === 'gemini-cli' || e.platform === 'gemini');
+  return summarizeOutcomeEntries(entries);
+}
+
+function formatGeminiFeedback(dir) {
+  const summary = getGeminiFeedbackSummary(dir);
+  if (!summary || Object.keys(summary).length === 0) {
+    return 'No Gemini feedback recorded yet. Use `npx nerviq --platform gemini feedback` to rate recommendations.';
+  }
+
+  const lines = ['Gemini Recommendation Feedback:', ''];
+  const entries = Array.isArray(summary) ? summary : Object.values(summary);
+  for (const entry of entries) {
+    const key = entry.key || 'unknown';
+    const accepted = entry.accepted || 0;
+    const rejected = entry.rejected || 0;
+    const total = entry.total || 0;
+    lines.push(`  ${key}: ${accepted} accepted, ${rejected} rejected (${total} total)`);
+  }
+  return lines.join('\n');
+}
+
+// --- Insights ---
+
+/**
+ * Generate Gemini-specific insights from audit history and feedback.
+ * Includes Gemini-unique patterns: sandbox-drift, policy-complexity,
+ * caveat-persistence, trust-class-stagnation.
+ */
+function generateGeminiInsights(dir) {
+  const history = getGeminiHistory(dir, 50);
+  const feedback = getGeminiFeedbackSummary(dir);
+  const insights = [];
+
+  // Pattern 1: Persistent failures
+  if (history.length >= 3) {
+    const recentFailKeys = new Map();
+    for (const entry of history.slice(0, 5)) {
+      for (const key of (entry.summary?.topActionKeys || [])) {
+        recentFailKeys.set(key, (recentFailKeys.get(key) || 0) + 1);
+      }
+    }
+    for (const [key, count] of recentFailKeys) {
+      if (count >= 3) {
+        insights.push({
+          type: 'persistent-failure',
+          severity: 'high',
+          key,
+          message: `Check ${key} has failed in ${count} of the last ${Math.min(history.length, 5)} audits. Consider addressing it or marking it as intentionally skipped.`,
+        });
+      }
+    }
+  }
+
+  // Pattern 2: Score regression
+  if (history.length >= 2) {
+    const scores = history.map(e => e.summary?.score ?? 0);
+    if (scores[0] < scores[1]) {
+      insights.push({
+        type: 'regression-pattern',
+        severity: 'medium',
+        message: `Score dropped from ${scores[1]} to ${scores[0]} in the most recent audit. Review recent changes.`,
+        delta: scores[0] - scores[1],
+      });
+    }
+  }
+
+  // Pattern 3: Improvement velocity stall
+  if (history.length >= 5) {
+    const recentScores = history.slice(0, 5).map(e => e.summary?.score ?? 0);
+    const range = Math.max(...recentScores) - Math.min(...recentScores);
+    if (range <= 2) {
+      insights.push({
+        type: 'velocity-stall',
+        severity: 'low',
+        message: `Score has been flat (range: ${range} points) over the last 5 audits. Consider addressing lower-priority checks.`,
+      });
+    }
+  }
+
+  // Pattern 4: Feedback signals
+  const feedbackEntries = Array.isArray(feedback) ? feedback : Object.values(feedback || {});
+  for (const entry of feedbackEntries) {
+    if (entry.rejected > entry.accepted && entry.total >= 2) {
+      insights.push({
+        type: 'feedback-signal',
+        severity: 'medium',
+        key: entry.key,
+        message: `Recommendation ${entry.key} has been rejected more than accepted (${entry.rejected}/${entry.total}). Consider adjusting or removing this recommendation.`,
+      });
+    }
+  }
+
+  // Gemini-specific Pattern 5: Sandbox drift
+  if (history.length >= 3) {
+    const sandboxKeys = [];
+    for (const entry of history.slice(0, 5)) {
+      for (const key of (entry.summary?.topActionKeys || [])) {
+        if (key.includes('sandbox') || key.includes('isolation')) {
+          sandboxKeys.push(key);
+        }
+      }
+    }
+    if (sandboxKeys.length >= 2) {
+      insights.push({
+        type: 'sandbox-drift',
+        severity: 'high',
+        message: `Sandbox-related checks have appeared in ${sandboxKeys.length} recent audits. Gemini sandbox configuration may be drifting from recommended posture.`,
+        keys: [...new Set(sandboxKeys)],
+      });
+    }
+  }
+
+  // Gemini-specific Pattern 6: Policy complexity
+  if (history.length >= 2) {
+    const latest = history[0];
+    const policyKeys = (latest.summary?.topActionKeys || []).filter(
+      k => k.includes('policy') || k.includes('governance') || k.includes('rule')
+    );
+    if (policyKeys.length >= 3) {
+      insights.push({
+        type: 'policy-complexity',
+        severity: 'medium',
+        message: `${policyKeys.length} policy/governance checks are failing simultaneously. Gemini policy configuration may be overly complex or misconfigured.`,
+        keys: policyKeys,
+      });
+    }
+  }
+
+  // Gemini-specific Pattern 7: Caveat persistence
+  if (history.length >= 3) {
+    const caveatKeys = new Map();
+    for (const entry of history.slice(0, 5)) {
+      for (const key of (entry.summary?.topActionKeys || [])) {
+        if (key.includes('caveat') || key.includes('warning') || key.includes('limitation')) {
+          caveatKeys.set(key, (caveatKeys.get(key) || 0) + 1);
+        }
+      }
+    }
+    for (const [key, count] of caveatKeys) {
+      if (count >= 3) {
+        insights.push({
+          type: 'caveat-persistence',
+          severity: 'medium',
+          key,
+          message: `Caveat/limitation ${key} has persisted across ${count} audits. This may indicate a Gemini CLI limitation that should be documented or worked around.`,
+        });
+      }
+    }
+  }
+
+  // Gemini-specific Pattern 8: Trust class stagnation
+  if (history.length >= 5) {
+    const trustKeys = new Map();
+    for (const entry of history.slice(0, 5)) {
+      for (const key of (entry.summary?.topActionKeys || [])) {
+        if (key.includes('trust') || key.includes('approval') || key.includes('permission')) {
+          trustKeys.set(key, (trustKeys.get(key) || 0) + 1);
+        }
+      }
+    }
+    for (const [key, count] of trustKeys) {
+      if (count >= 4) {
+        insights.push({
+          type: 'trust-class-stagnation',
+          severity: 'medium',
+          key,
+          message: `Trust/permission check ${key} has been failing in ${count} of the last 5 audits. Gemini trust classification may need elevation or explicit override.`,
+        });
+      }
+    }
+  }
+
+  return {
+    platform: 'gemini',
+    generatedAt: new Date().toISOString(),
+    snapshotCount: history.length,
+    feedbackCount: feedbackEntries.length,
+    insights,
+    summary: insights.length === 0
+      ? 'No actionable insights detected. Keep running audits to build pattern data.'
+      : `${insights.length} insight(s) detected across ${history.length} snapshots.`,
+  };
+}
+
+function formatGeminiInsights(dir) {
+  const result = generateGeminiInsights(dir);
+  if (result.insights.length === 0) {
+    return result.summary;
+  }
+
+  const lines = ['Gemini Insights:', ''];
+  for (const insight of result.insights) {
+    const severity = insight.severity.toUpperCase();
+    lines.push(`  [${severity}] ${insight.message}`);
+  }
+  lines.push('');
+  lines.push(result.summary);
+  return lines.join('\n');
+}
+
+module.exports = {
+  getGeminiHistory,
+  formatGeminiHistory,
+  compareGeminiLatest,
+  exportGeminiTrendReport,
+  recordGeminiFeedback,
+  getGeminiFeedbackSummary,
+  formatGeminiFeedback,
+  generateGeminiInsights,
+  formatGeminiInsights,
+};