@rune-kit/rune 2.8.0 → 2.11.0

package/extensions/analytics/PACK.md

@@ -1,92 +1,92 @@
----
-name: "@rune/analytics"
-description: Analytics patterns — tracking setup, A/B testing, funnel analysis, and dashboard design.
-metadata:
-  author: runedev
-  version: "0.2.0"
-  layer: L4
-  price: "$12"
-  target: Growth engineers
-  format: split
----
-
-# @rune/analytics
-
-## Purpose
-
-Analytics implementations fail silently: tracking events that fire but never reach the dashboard because the event name has a typo, A/B tests that run for weeks without reaching statistical significance because the sample size was never calculated, funnel reports that show a 90% drop-off that's actually a tracking gap, and dashboards that load 500K rows client-side because the aggregation happens in the browser instead of the database. This pack covers the full analytics stack — instrumentation, experimentation, analysis, and visualization — with patterns that produce data you can actually trust and act on.
-
-## Triggers
-
-- Auto-trigger: when `gtag`, `posthog`, `mixpanel`, `plausible`, `analytics`, `experiment`, `feature-flag`, `launchdarkly` detected
-- `/rune tracking-setup` — set up or audit analytics tracking
-- `/rune ab-testing` — design and implement A/B experiments
-- `/rune funnel-analysis` — build conversion funnel tracking
-- `/rune dashboard-patterns` — build analytics dashboard
-- Called by `cook` (L1) when analytics feature requested
-- Called by `marketing` (L2) when measuring campaign performance
-
-## Skills Included
-
-| Skill | Model | Description |
-|-------|-------|-------------|
-| [tracking-setup](skills/tracking-setup.md) | sonnet | GA4, Plausible, PostHog, Mixpanel — event taxonomy design, consent management, server-side tracking, UTM handling. |
-| [ab-testing](skills/ab-testing.md) | sonnet | Experiment design, statistical significance, feature flags (LaunchDarkly, Unleash), rollout strategies, result analysis. |
-| [funnel-analysis](skills/funnel-analysis.md) | sonnet | Conversion tracking, drop-off identification, cohort analysis, retention metrics, LTV calculation, attribution modeling. |
-| [dashboard-patterns](skills/dashboard-patterns.md) | sonnet | KPI cards, time series charts, comparison views, drill-down navigation, export functionality, real-time counters. |
-| [sql-patterns](skills/sql-patterns.md) | sonnet | Aggregations, window functions, CTEs, performance optimization, and safe parameterized queries for analytics workloads. |
-| [data-validation](skills/data-validation.md) | sonnet | Input validation, schema enforcement, data pipeline checks, anomaly detection, and data freshness monitoring. |
-| [statistical-analysis](skills/statistical-analysis.md) | sonnet | Significance testing, regression basics, distribution analysis, and correlation detection for product metrics. |
-
-## Tech Stack Support
-
-| Area | Options | Notes |
-|------|---------|-------|
-| Analytics | GA4, Plausible, PostHog, Mixpanel | Plausible for privacy-first; PostHog for product analytics |
-| Feature Flags | LaunchDarkly, Unleash, GrowthBook | GrowthBook open-source with built-in A/B |
-| Charts | Recharts, Tremor, Chart.js, D3 | Tremor best for dashboards; D3 for custom visualizations |
-| Database | PostgreSQL + aggregation views | Pre-aggregate for dashboard performance |
-
-## Connections
-
-```
-Calls → @rune/ui (L4): dashboard components
-Calls → @rune/backend (L4): tracking API setup
-Called By ← marketing (L2): measuring campaign performance
-Called By ← cook (L1): when analytics feature requested
-```
-
-## Constraints
-
-1. MUST use typed event taxonomy — ad-hoc event names create unmaintainable analytics that nobody trusts.
-2. MUST implement consent management before any tracking — GDPR/CCPA compliance is non-negotiable.
-3. MUST calculate sample size before starting A/B tests — running experiments without power analysis wastes time and produces meaningless results.
-4. MUST aggregate data server-side for dashboards — sending raw events to the client causes slow loads and exposes user data.
-5. MUST persist variant assignment per user — inconsistent assignment invalidates experiment results.
-
-## Sharp Edges
-
-| Failure Mode | Severity | Mitigation |
-|---|---|---|
-| Peeking at A/B test results before reaching sample size (false positive) | HIGH | Lock results until sample size reached; show "not yet significant" warning |
-| Event name typo means data goes to wrong metric (silent data loss) | HIGH | Typed event taxonomy with TypeScript union; no raw string event names |
-| Ad blockers drop 30-40% of client-side tracking events | HIGH | Implement server-side tracking proxy (`/api/analytics`); use `sendBeacon` |
-| Dashboard loads 500K raw events client-side (browser freezes) | HIGH | Pre-aggregate in SQL; paginate time series; lazy-load off-screen charts |
-| Same user gets different A/B variant across sessions (polluted results) | MEDIUM | Hash user ID + experiment ID for deterministic assignment; persist in cookie |
-| Funnel shows 0% conversion because step events use different flow IDs | MEDIUM | Generate flow ID at funnel entry; pass through all steps; validate correlation |
-
-## Done When
-
-- Event tracking fires with typed taxonomy and consent management
-- A/B testing assigns persistent variants with sample size calculation
-- Funnel analysis tracks correlated steps with drop-off rates
-- Dashboard renders KPI cards with comparison, time series, and export
-- Server-side tracking proxy handles ad-blocked clients
-- SQL queries use parameterized statements, proper indexing, and cursor-based pagination
-- Data pipeline validates inputs with schema enforcement and anomaly detection
-- Statistical tests applied correctly (right method for right question)
-- Structured report emitted for each skill invoked
-
-## Cost Profile
-
-~8,000–14,000 tokens per full pack run (all 7 skills). Individual skill: ~2,000–4,000 tokens. Sonnet default. Use haiku for detection scans; escalate to sonnet for experiment design and dashboard patterns.
+---
+name: "@rune/analytics"
+description: Analytics patterns — tracking setup, A/B testing, funnel analysis, and dashboard design.
+metadata:
+  author: runedev
+  version: "0.2.0"
+  layer: L4
+  price: "$12"
+  target: Growth engineers
+  format: split
+---
+
+# @rune/analytics
+
+## Purpose
+
+Analytics implementations fail silently: tracking events that fire but never reach the dashboard because the event name has a typo, A/B tests that run for weeks without reaching statistical significance because the sample size was never calculated, funnel reports that show a 90% drop-off that's actually a tracking gap, and dashboards that load 500K rows client-side because the aggregation happens in the browser instead of the database. This pack covers the full analytics stack — instrumentation, experimentation, analysis, and visualization — with patterns that produce data you can actually trust and act on.
+
+## Triggers
+
+- Auto-trigger: when `gtag`, `posthog`, `mixpanel`, `plausible`, `analytics`, `experiment`, `feature-flag`, `launchdarkly` detected
+- `/rune tracking-setup` — set up or audit analytics tracking
+- `/rune ab-testing` — design and implement A/B experiments
+- `/rune funnel-analysis` — build conversion funnel tracking
+- `/rune dashboard-patterns` — build analytics dashboard
+- Called by `cook` (L1) when analytics feature requested
+- Called by `marketing` (L2) when measuring campaign performance
+
+## Skills Included
+
+| Skill | Model | Description |
+|-------|-------|-------------|
+| [tracking-setup](skills/tracking-setup.md) | sonnet | GA4, Plausible, PostHog, Mixpanel — event taxonomy design, consent management, server-side tracking, UTM handling. |
+| [ab-testing](skills/ab-testing.md) | sonnet | Experiment design, statistical significance, feature flags (LaunchDarkly, Unleash), rollout strategies, result analysis. |
+| [funnel-analysis](skills/funnel-analysis.md) | sonnet | Conversion tracking, drop-off identification, cohort analysis, retention metrics, LTV calculation, attribution modeling. |
+| [dashboard-patterns](skills/dashboard-patterns.md) | sonnet | KPI cards, time series charts, comparison views, drill-down navigation, export functionality, real-time counters. |
+| [sql-patterns](skills/sql-patterns.md) | sonnet | Aggregations, window functions, CTEs, performance optimization, and safe parameterized queries for analytics workloads. |
+| [data-validation](skills/data-validation.md) | sonnet | Input validation, schema enforcement, data pipeline checks, anomaly detection, and data freshness monitoring. |
+| [statistical-analysis](skills/statistical-analysis.md) | sonnet | Significance testing, regression basics, distribution analysis, and correlation detection for product metrics. |
+
+## Tech Stack Support
+
+| Area | Options | Notes |
+|------|---------|-------|
+| Analytics | GA4, Plausible, PostHog, Mixpanel | Plausible for privacy-first; PostHog for product analytics |
+| Feature Flags | LaunchDarkly, Unleash, GrowthBook | GrowthBook open-source with built-in A/B |
+| Charts | Recharts, Tremor, Chart.js, D3 | Tremor best for dashboards; D3 for custom visualizations |
+| Database | PostgreSQL + aggregation views | Pre-aggregate for dashboard performance |
+
+## Connections
+
+```
+Calls → @rune/ui (L4): dashboard components
+Calls → @rune/backend (L4): tracking API setup
+Called By ← marketing (L2): measuring campaign performance
+Called By ← cook (L1): when analytics feature requested
+```
+
+## Constraints
+
+1. MUST use typed event taxonomy — ad-hoc event names create unmaintainable analytics that nobody trusts.
+2. MUST implement consent management before any tracking — GDPR/CCPA compliance is non-negotiable.
+3. MUST calculate sample size before starting A/B tests — running experiments without power analysis wastes time and produces meaningless results.
+4. MUST aggregate data server-side for dashboards — sending raw events to the client causes slow loads and exposes user data.
+5. MUST persist variant assignment per user — inconsistent assignment invalidates experiment results.
+
+## Sharp Edges
+
+| Failure Mode | Severity | Mitigation |
+|---|---|---|
+| Peeking at A/B test results before reaching sample size (false positive) | HIGH | Lock results until sample size reached; show "not yet significant" warning |
+| Event name typo means data goes to wrong metric (silent data loss) | HIGH | Typed event taxonomy with TypeScript union; no raw string event names |
+| Ad blockers drop 30-40% of client-side tracking events | HIGH | Implement server-side tracking proxy (`/api/analytics`); use `sendBeacon` |
+| Dashboard loads 500K raw events client-side (browser freezes) | HIGH | Pre-aggregate in SQL; paginate time series; lazy-load off-screen charts |
+| Same user gets different A/B variant across sessions (polluted results) | MEDIUM | Hash user ID + experiment ID for deterministic assignment; persist in cookie |
+| Funnel shows 0% conversion because step events use different flow IDs | MEDIUM | Generate flow ID at funnel entry; pass through all steps; validate correlation |
+
+## Done When
+
+- Event tracking fires with typed taxonomy and consent management
+- A/B testing assigns persistent variants with sample size calculation
+- Funnel analysis tracks correlated steps with drop-off rates
+- Dashboard renders KPI cards with comparison, time series, and export
+- Server-side tracking proxy handles ad-blocked clients
+- SQL queries use parameterized statements, proper indexing, and cursor-based pagination
+- Data pipeline validates inputs with schema enforcement and anomaly detection
+- Statistical tests applied correctly (right method for right question)
+- Structured report emitted for each skill invoked
+
+## Cost Profile
+
+~8,000–14,000 tokens per full pack run (all 7 skills). Individual skill: ~2,000–4,000 tokens. Sonnet default. Use haiku for detection scans; escalate to sonnet for experiment design and dashboard patterns.

package/extensions/analytics/skills/ab-testing.md

@@ -1,72 +1,72 @@
----
-name: "ab-testing"
-pack: "@rune/analytics"
-description: "A/B testing patterns — experiment design, statistical significance, feature flags (LaunchDarkly, Unleash), rollout strategies, result analysis."
-model: sonnet
-tools: [Read, Edit, Write, Grep, Glob, Bash]
----
-
-# ab-testing
-
-A/B testing patterns — experiment design, statistical significance, feature flags (LaunchDarkly, Unleash), rollout strategies, result analysis.
-
-#### Workflow
-
-**Step 1 — Detect experiment setup**
-Use Grep to find experiment code: `useFeatureFlag`, `useExperiment`, `LaunchDarkly`, `Unleash`, `GrowthBook`, `variant`, `experiment`. Read feature flag initialization and variant assignment to understand: flag provider, assignment method (random, user-based, percentage), and metric collection.
-
-**Step 2 — Audit experiment validity**
-Check for: no sample size calculation (experiment runs indefinitely), peeking at results before significance (inflated false positive rate), no control group definition, variant assignment not persisted across sessions (same user sees different variants), metrics not tracked per-variant (can't measure impact), and feature flags without cleanup (dead flags accumulate).
-
-**Step 3 — Emit experiment patterns**
-Emit: experiment setup with sample size calculator, persistent variant assignment (cookie/user-ID based), metric collection per variant, significance calculator, and feature flag lifecycle with cleanup reminder.
-
-#### Example
-
-```typescript
-// A/B experiment with persistent assignment and significance check
-import { z } from 'zod';
-
-const ExperimentSchema = z.object({
-  id: z.string(),
-  variants: z.array(z.object({ id: z.string(), weight: z.number() })),
-  metrics: z.array(z.string()),
-});
-
-// Persistent variant assignment (deterministic hash)
-function assignVariant(userId: string, experimentId: string, variants: { id: string; weight: number }[]): string {
-  const hash = cyrb53(`${userId}:${experimentId}`);
-  const normalized = (hash % 10000) / 10000; // [0, 1)
-  let cumulative = 0;
-  for (const variant of variants) {
-    cumulative += variant.weight;
-    if (normalized < cumulative) return variant.id;
-  }
-  return variants[variants.length - 1].id;
-}
-
-// Simple hash function (deterministic, fast)
-function cyrb53(str: string): number {
-  let h1 = 0xdeadbeef, h2 = 0x41c6ce57;
-  for (let i = 0; i < str.length; i++) {
-    const ch = str.charCodeAt(i);
-    h1 = Math.imul(h1 ^ ch, 2654435761);
-    h2 = Math.imul(h2 ^ ch, 1597334677);
-  }
-  h1 = Math.imul(h1 ^ (h1 >>> 16), 2246822507);
-  h2 = Math.imul(h2 ^ (h2 >>> 13), 3266489909);
-  return 4294967296 * (2097151 & h2) + (h1 >>> 0);
-}
-
-// Sample size calculator (two-proportion z-test)
-function requiredSampleSize(baselineRate: number, mde: number, power = 0.8, alpha = 0.05): number {
-  const zAlpha = 1.96; // alpha=0.05 two-tailed
-  const zBeta = 0.842; // power=0.8
-  const p1 = baselineRate;
-  const p2 = baselineRate * (1 + mde);
-  const pooled = (p1 + p2) / 2;
-  return Math.ceil(
-    (2 * pooled * (1 - pooled) * Math.pow(zAlpha + zBeta, 2)) / Math.pow(p2 - p1, 2),
-  );
-}
-```
+---
+name: "ab-testing"
+pack: "@rune/analytics"
+description: "A/B testing patterns — experiment design, statistical significance, feature flags (LaunchDarkly, Unleash), rollout strategies, result analysis."
+model: sonnet
+tools: [Read, Edit, Write, Grep, Glob, Bash]
+---
+
+# ab-testing
+
+A/B testing patterns — experiment design, statistical significance, feature flags (LaunchDarkly, Unleash), rollout strategies, result analysis.
+
+#### Workflow
+
+**Step 1 — Detect experiment setup**
+Use Grep to find experiment code: `useFeatureFlag`, `useExperiment`, `LaunchDarkly`, `Unleash`, `GrowthBook`, `variant`, `experiment`. Read feature flag initialization and variant assignment to understand: flag provider, assignment method (random, user-based, percentage), and metric collection.
+
+**Step 2 — Audit experiment validity**
+Check for: no sample size calculation (experiment runs indefinitely), peeking at results before significance (inflated false positive rate), no control group definition, variant assignment not persisted across sessions (same user sees different variants), metrics not tracked per-variant (can't measure impact), and feature flags without cleanup (dead flags accumulate).
+
+**Step 3 — Emit experiment patterns**
+Emit: experiment setup with sample size calculator, persistent variant assignment (cookie/user-ID based), metric collection per variant, significance calculator, and feature flag lifecycle with cleanup reminder.
+
+#### Example
+
+```typescript
+// A/B experiment with persistent assignment and significance check
+import { z } from 'zod';
+
+const ExperimentSchema = z.object({
+  id: z.string(),
+  variants: z.array(z.object({ id: z.string(), weight: z.number() })),
+  metrics: z.array(z.string()),
+});
+
+// Persistent variant assignment (deterministic hash)
+function assignVariant(userId: string, experimentId: string, variants: { id: string; weight: number }[]): string {
+  const hash = cyrb53(`${userId}:${experimentId}`);
+  const normalized = (hash % 10000) / 10000; // [0, 1)
+  let cumulative = 0;
+  for (const variant of variants) {
+    cumulative += variant.weight;
+    if (normalized < cumulative) return variant.id;
+  }
+  return variants[variants.length - 1].id;
+}
+
+// Simple hash function (deterministic, fast)
+function cyrb53(str: string): number {
+  let h1 = 0xdeadbeef, h2 = 0x41c6ce57;
+  for (let i = 0; i < str.length; i++) {
+    const ch = str.charCodeAt(i);
+    h1 = Math.imul(h1 ^ ch, 2654435761);
+    h2 = Math.imul(h2 ^ ch, 1597334677);
+  }
+  h1 = Math.imul(h1 ^ (h1 >>> 16), 2246822507);
+  h2 = Math.imul(h2 ^ (h2 >>> 13), 3266489909);
+  return 4294967296 * (2097151 & h2) + (h1 >>> 0);
+}
+
+// Sample size calculator (two-proportion z-test)
+function requiredSampleSize(baselineRate: number, mde: number, power = 0.8, alpha = 0.05): number {
+  const zAlpha = 1.96; // alpha=0.05 two-tailed
+  const zBeta = 0.842; // power=0.8
+  const p1 = baselineRate;
+  const p2 = baselineRate * (1 + mde);
+  const pooled = (p1 + p2) / 2;
+  return Math.ceil(
+    (2 * pooled * (1 - pooled) * Math.pow(zAlpha + zBeta, 2)) / Math.pow(p2 - p1, 2),
+  );
+}
+```

package/extensions/analytics/skills/dashboard-patterns.md

@@ -1,83 +1,83 @@
----
-name: "dashboard-patterns"
-pack: "@rune/analytics"
-description: "Analytics dashboard design — KPI cards, time series charts, comparison views, drill-down navigation, export functionality, real-time counters."
-model: sonnet
-tools: [Read, Edit, Write, Grep, Glob, Bash]
----
-
-# dashboard-patterns
-
-Analytics dashboard design — KPI cards, time series charts, comparison views, drill-down navigation, export functionality, real-time counters.
-
-#### Workflow
-
-**Step 1 — Detect dashboard components**
-Use Grep to find dashboard code: `Chart`, `recharts`, `chart.js`, `d3`, `tremor`, `KPI`, `metric`, `dashboard`. Read dashboard pages and data fetching to understand: charting library, data source (API, database, analytics provider), refresh strategy, and component structure.
-
-**Step 2 — Audit dashboard performance**
-Check for: all data fetched on page load (no lazy loading for off-screen charts), no time range selector (stuck on one period), raw data sent to client for aggregation (should aggregate server-side), no loading states (charts pop in), missing comparison period (no "vs last week"), no data export, and charts re-rendering on unrelated state changes.
-
-**Step 3 — Emit dashboard patterns**
-Emit: KPI card with comparison indicator, time series chart with range selector, server-side aggregation endpoint, lazy-loaded chart sections, and CSV export utility.
-
-#### Example
-
-```tsx
-// Dashboard KPI card with comparison
-interface KpiProps {
-  label: string;
-  value: number;
-  previousValue: number;
-  format: 'number' | 'currency' | 'percent';
-}
-
-function KpiCard({ label, value, previousValue, format }: KpiProps) {
-  const change = previousValue ? ((value - previousValue) / previousValue) * 100 : 0;
-  const formatted = format === 'currency'
-    ? new Intl.NumberFormat('en-US', { style: 'currency', currency: 'USD', maximumFractionDigits: 0 }).format(value)
-    : format === 'percent'
-    ? `${value.toFixed(1)}%`
-    : new Intl.NumberFormat('en-US', { notation: 'compact' }).format(value);
-
-  return (
-    <div className="rounded-lg border bg-card p-6">
-      <p className="text-sm text-muted-foreground">{label}</p>
-      <p className="text-2xl font-bold font-mono mt-1">{formatted}</p>
-      <p className={`text-sm mt-1 ${change >= 0 ? 'text-green-600' : 'text-red-600'}`}>
-        {change >= 0 ? '▲' : '▼'} {Math.abs(change).toFixed(1)}% vs previous period
-      </p>
-    </div>
-  );
-}
-
-// Server-side aggregation endpoint — app/api/metrics/route.ts
-export async function GET(req: Request) {
-  const { searchParams } = new URL(req.url);
-  const range = searchParams.get('range') || '7d';
-  const interval = range === '24h' ? 'hour' : range === '7d' ? 'day' : 'week';
-
-  const metrics = await db.execute(sql`
-    SELECT DATE_TRUNC(${interval}, timestamp) AS period,
-      COUNT(*) AS page_views,
-      COUNT(DISTINCT user_id) AS unique_visitors,
-      COUNT(*) FILTER (WHERE name = 'signup_completed') AS signups
-    FROM events
-    WHERE timestamp > NOW() - ${range}::interval
-    GROUP BY period ORDER BY period
-  `);
-  return Response.json(metrics);
-}
-
-// CSV export utility
-function exportCsv(data: Record<string, unknown>[], filename: string) {
-  const headers = Object.keys(data[0]);
-  const csv = [headers.join(','), ...data.map(row => headers.map(h => JSON.stringify(row[h] ?? '')).join(','))].join('\n');
-  const blob = new Blob([csv], { type: 'text/csv' });
-  const a = document.createElement('a');
-  a.href = URL.createObjectURL(blob);
-  a.download = `${filename}-${new Date().toISOString().split('T')[0]}.csv`;
-  a.click();
-  URL.revokeObjectURL(a.href);
-}
-```
+---
+name: "dashboard-patterns"
+pack: "@rune/analytics"
+description: "Analytics dashboard design — KPI cards, time series charts, comparison views, drill-down navigation, export functionality, real-time counters."
+model: sonnet
+tools: [Read, Edit, Write, Grep, Glob, Bash]
+---
+
+# dashboard-patterns
+
+Analytics dashboard design — KPI cards, time series charts, comparison views, drill-down navigation, export functionality, real-time counters.
+
+#### Workflow
+
+**Step 1 — Detect dashboard components**
+Use Grep to find dashboard code: `Chart`, `recharts`, `chart.js`, `d3`, `tremor`, `KPI`, `metric`, `dashboard`. Read dashboard pages and data fetching to understand: charting library, data source (API, database, analytics provider), refresh strategy, and component structure.
+
+**Step 2 — Audit dashboard performance**
+Check for: all data fetched on page load (no lazy loading for off-screen charts), no time range selector (stuck on one period), raw data sent to client for aggregation (should aggregate server-side), no loading states (charts pop in), missing comparison period (no "vs last week"), no data export, and charts re-rendering on unrelated state changes.
+
+**Step 3 — Emit dashboard patterns**
+Emit: KPI card with comparison indicator, time series chart with range selector, server-side aggregation endpoint, lazy-loaded chart sections, and CSV export utility.
+
+#### Example
+
+```tsx
+// Dashboard KPI card with comparison
+interface KpiProps {
+  label: string;
+  value: number;
+  previousValue: number;
+  format: 'number' | 'currency' | 'percent';
+}
+
+function KpiCard({ label, value, previousValue, format }: KpiProps) {
+  const change = previousValue ? ((value - previousValue) / previousValue) * 100 : 0;
+  const formatted = format === 'currency'
+    ? new Intl.NumberFormat('en-US', { style: 'currency', currency: 'USD', maximumFractionDigits: 0 }).format(value)
+    : format === 'percent'
+    ? `${value.toFixed(1)}%`
+    : new Intl.NumberFormat('en-US', { notation: 'compact' }).format(value);
+
+  return (
+    <div className="rounded-lg border bg-card p-6">
+      <p className="text-sm text-muted-foreground">{label}</p>
+      <p className="text-2xl font-bold font-mono mt-1">{formatted}</p>
+      <p className={`text-sm mt-1 ${change >= 0 ? 'text-green-600' : 'text-red-600'}`}>
+        {change >= 0 ? '▲' : '▼'} {Math.abs(change).toFixed(1)}% vs previous period
+      </p>
+    </div>
+  );
+}
+
+// Server-side aggregation endpoint — app/api/metrics/route.ts
+export async function GET(req: Request) {
+  const { searchParams } = new URL(req.url);
+  const range = searchParams.get('range') || '7d';
+  const interval = range === '24h' ? 'hour' : range === '7d' ? 'day' : 'week';
+
+  const metrics = await db.execute(sql`
+    SELECT DATE_TRUNC(${interval}, timestamp) AS period,
+      COUNT(*) AS page_views,
+      COUNT(DISTINCT user_id) AS unique_visitors,
+      COUNT(*) FILTER (WHERE name = 'signup_completed') AS signups
+    FROM events
+    WHERE timestamp > NOW() - ${range}::interval
+    GROUP BY period ORDER BY period
+  `);
+  return Response.json(metrics);
+}
+
+// CSV export utility
+function exportCsv(data: Record<string, unknown>[], filename: string) {
+  const headers = Object.keys(data[0]);
+  const csv = [headers.join(','), ...data.map(row => headers.map(h => JSON.stringify(row[h] ?? '')).join(','))].join('\n');
+  const blob = new Blob([csv], { type: 'text/csv' });
+  const a = document.createElement('a');
+  a.href = URL.createObjectURL(blob);
+  a.download = `${filename}-${new Date().toISOString().split('T')[0]}.csv`;
+  a.click();
+  URL.revokeObjectURL(a.href);
+}
+```

package/extensions/analytics/skills/data-validation.md

@@ -1,68 +1,68 @@
----
-name: "data-validation"
-pack: "@rune/analytics"
-description: "Data quality patterns — input validation, schema enforcement, data pipeline checks, anomaly detection, and data freshness monitoring."
-model: sonnet
-tools: [Read, Edit, Write, Grep, Glob, Bash]
----
-
-# data-validation
-
-Data quality patterns — input validation, schema enforcement, data pipeline checks, anomaly detection, and data freshness monitoring.
-
-#### Workflow
-
-**Step 1 — Detect data flows**
-Use Grep to find data ingestion points: API endpoints that accept data, CSV/JSON import handlers, webhook receivers, database seed scripts, ETL pipelines. Map: source → transform → destination for each flow.
-
-**Step 2 — Audit data quality**
-Check for: missing input validation on data ingestion endpoints, no schema validation on imported files, no null/empty checks on required fields, no data type coercion (string "123" stored as string not number), no anomaly detection (sudden 10x spike in values), no data freshness check ("when was this data last updated?"), and no deduplication on event streams.
-
-**Step 3 — Emit validation patterns**
-Emit: schema validation with Zod for API inputs, data pipeline validation middleware, anomaly detection query, data freshness monitor, and deduplication patterns.
-
-#### Example
-
-```typescript
-import { z } from 'zod';
-
-// Data pipeline validation schema
-const MetricRowSchema = z.object({
-  timestamp: z.coerce.date(),
-  metric_name: z.string().min(1).max(100),
-  value: z.number().finite(),
-  source: z.enum(['api', 'webhook', 'import', 'manual']),
-  tags: z.record(z.string()).optional(),
-});
-
-// Batch validation with error collection (not fail-fast)
-function validateBatch(rows: unknown[]): { valid: z.infer<typeof MetricRowSchema>[]; errors: { row: number; error: string }[] } {
-  const valid: z.infer<typeof MetricRowSchema>[] = [];
-  const errors: { row: number; error: string }[] = [];
-  rows.forEach((row, i) => {
-    const result = MetricRowSchema.safeParse(row);
-    if (result.success) valid.push(result.data);
-    else errors.push({ row: i, error: result.error.issues.map(e => e.message).join('; ') });
-  });
-  return { valid, errors };
-}
-
-// Anomaly detection — flag values >3 standard deviations from rolling mean
-// SELECT metric_name, value, timestamp,
-//   AVG(value) OVER (PARTITION BY metric_name ORDER BY timestamp ROWS 30 PRECEDING) AS rolling_mean,
-//   STDDEV(value) OVER (PARTITION BY metric_name ORDER BY timestamp ROWS 30 PRECEDING) AS rolling_std
-// FROM metrics
-// HAVING ABS(value - rolling_mean) > 3 * rolling_std;
-
-// Data freshness monitor
-async function checkFreshness(tables: string[], maxStaleMinutes: number) {
-  const stale: string[] = [];
-  for (const table of tables) {
-    const result = await db.query(
-      `SELECT EXTRACT(EPOCH FROM NOW() - MAX(updated_at)) / 60 AS minutes_stale FROM ${table}`
-    );
-    if (result.rows[0]?.minutes_stale > maxStaleMinutes) stale.push(table);
-  }
-  return stale;
-}
-```
+---
+name: "data-validation"
+pack: "@rune/analytics"
+description: "Data quality patterns — input validation, schema enforcement, data pipeline checks, anomaly detection, and data freshness monitoring."
+model: sonnet
+tools: [Read, Edit, Write, Grep, Glob, Bash]
+---
+
+# data-validation
+
+Data quality patterns — input validation, schema enforcement, data pipeline checks, anomaly detection, and data freshness monitoring.
+
+#### Workflow
+
+**Step 1 — Detect data flows**
+Use Grep to find data ingestion points: API endpoints that accept data, CSV/JSON import handlers, webhook receivers, database seed scripts, ETL pipelines. Map: source → transform → destination for each flow.
+
+**Step 2 — Audit data quality**
+Check for: missing input validation on data ingestion endpoints, no schema validation on imported files, no null/empty checks on required fields, no data type coercion (string "123" stored as string not number), no anomaly detection (sudden 10x spike in values), no data freshness check ("when was this data last updated?"), and no deduplication on event streams.
+
+**Step 3 — Emit validation patterns**
+Emit: schema validation with Zod for API inputs, data pipeline validation middleware, anomaly detection query, data freshness monitor, and deduplication patterns.
+
+#### Example
+
+```typescript
+import { z } from 'zod';
+
+// Data pipeline validation schema
+const MetricRowSchema = z.object({
+  timestamp: z.coerce.date(),
+  metric_name: z.string().min(1).max(100),
+  value: z.number().finite(),
+  source: z.enum(['api', 'webhook', 'import', 'manual']),
+  tags: z.record(z.string()).optional(),
+});
+
+// Batch validation with error collection (not fail-fast)
+function validateBatch(rows: unknown[]): { valid: z.infer<typeof MetricRowSchema>[]; errors: { row: number; error: string }[] } {
+  const valid: z.infer<typeof MetricRowSchema>[] = [];
+  const errors: { row: number; error: string }[] = [];
+  rows.forEach((row, i) => {
+    const result = MetricRowSchema.safeParse(row);
+    if (result.success) valid.push(result.data);
+    else errors.push({ row: i, error: result.error.issues.map(e => e.message).join('; ') });
+  });
+  return { valid, errors };
+}
+
+// Anomaly detection — flag values >3 standard deviations from rolling mean
+// SELECT metric_name, value, timestamp,
+//   AVG(value) OVER (PARTITION BY metric_name ORDER BY timestamp ROWS 30 PRECEDING) AS rolling_mean,
+//   STDDEV(value) OVER (PARTITION BY metric_name ORDER BY timestamp ROWS 30 PRECEDING) AS rolling_std
+// FROM metrics
+// HAVING ABS(value - rolling_mean) > 3 * rolling_std;
+
+// Data freshness monitor
+async function checkFreshness(tables: string[], maxStaleMinutes: number) {
+  const stale: string[] = [];
+  for (const table of tables) {
+    const result = await db.query(
+      `SELECT EXTRACT(EPOCH FROM NOW() - MAX(updated_at)) / 60 AS minutes_stale FROM ${table}`
+    );
+    if (result.rows[0]?.minutes_stale > maxStaleMinutes) stale.push(table);
+  }
+  return stale;
+}
+```