/impact

<objective> Connect code to its real-world impact.

Current AI has no concept of production reality. It generates code in a vacuum.

Reality-Grounded Development links:

• Code → Production metrics
• Features → Business outcomes
• Architecture → Infrastructure costs
• Code paths → Incident history

Decisions become informed by reality, not just code aesthetics. </objective>

<philosophy> Code exists to produce outcomes.

A function that runs 1M times/day matters more than one that runs once/month. A feature that drives 20% of revenue deserves more attention. An endpoint that caused 3 outages needs more care.

Without this context, all code looks equally important. With it, priorities become obvious. </philosophy>

<reference> See `.planning/config.json` for data source configuration and integration setup. </reference>

<data_sources>

Reality Data Sources

Production Metrics

• Request volumes per endpoint
• Error rates per function
• Latency percentiles
• Resource consumption

Integration points:

• Datadog, New Relic, Prometheus
• Application logs
• APM traces

Business Metrics

• Feature → Revenue correlation
• User engagement by feature
• Conversion impact
• Retention correlation

Integration points:

• Analytics platforms (Amplitude, Mixpanel)
• Business dashboards
• A/B test results

Cost Data

• Compute costs by service
• Storage costs by table
• Network costs by endpoint
• Third-party API costs

Integration points:

• AWS Cost Explorer
• Cloud billing APIs
• Vendor invoices

Incident History

• Outages linked to code changes
• Error spikes by file
• On-call pages by service
• Post-mortems by feature

Integration points:

• PagerDuty, Opsgenie
• Incident management systems
• Post-mortem databases </data_sources>

<commands> ## Subcommands

/gywd:impact <target>

Show impact profile for file, function, or feature.

## Impact Profile: src/api/checkout.ts

### Production Metrics
| Metric | Value | Trend |
|--------|-------|-------|
| Requests/day | 47,000 | ↑ 12% |
| Error rate | 0.3% | → stable |
| P50 latency | 120ms | → stable |
| P99 latency | 890ms | ↓ improved |

### Business Impact
- **Revenue attribution**: 34% of daily revenue
- **Conversion impact**: Critical path for purchase
- **User sessions**: 23% of all sessions touch this

### Cost Profile
- **Compute**: $340/month (2.1% of total)
- **Database**: 12% of read queries
- **External APIs**: $89/month (Stripe calls)

### Incident History
| Date | Severity | Cause | Duration |
|------|----------|-------|----------|
| 2024-01-15 | P1 | Race condition | 45 min |
| 2023-11-03 | P2 | Timeout spike | 20 min |

### Risk Assessment
🔴 **High impact surface**: Outage here affects revenue
🟠 **Moderate stability**: 2 incidents in 6 months
🟢 **Good performance**: Latency within SLO

### Recommendations
1. Add circuit breaker (high impact, no protection)
2. Increase test coverage (currently 62%)
3. Consider caching (high volume, stable data)

/gywd:impact --metrics

Show production metrics across codebase.

## Production Metrics Overview

### Top Traffic (Requests/Day)
| Endpoint | Requests | Errors | P99 |
|----------|----------|--------|-----|
| GET /api/products | 892K | 0.1% | 45ms |
| POST /api/cart | 234K | 0.2% | 120ms |
| POST /api/checkout | 47K | 0.3% | 890ms |
| GET /api/user | 1.2M | 0.05% | 30ms |

### Error Hotspots
| File | Error Rate | Volume | Trend |
|------|------------|--------|-------|
| src/services/payment.ts | 1.2% | 12K/day | ↑ |
| src/api/orders.ts | 0.8% | 89K/day | → |
| src/utils/shipping.ts | 0.6% | 34K/day | ↓ |

### Performance Concerns
| Endpoint | P99 | SLO | Status |
|----------|-----|-----|--------|
| POST /api/search | 2.3s | 1s | 🔴 Violation |
| GET /api/recommendations | 1.8s | 2s | 🟡 Warning |

/gywd:impact --cost

Show infrastructure costs by code area.

## Cost Attribution

**Total Monthly**: $12,400

### By Service
| Service | Monthly | % | Trend |
|---------|---------|---|-------|
| API Server | $4,200 | 34% | ↑ |
| Database | $3,100 | 25% | → |
| Cache (Redis) | $890 | 7% | → |
| Storage (S3) | $2,100 | 17% | ↑ |
| CDN | $1,200 | 10% | → |
| External APIs | $910 | 7% | ↓ |

### Cost Drivers
1. **Image processing** (S3 + compute): $2,800/month
   - Opportunity: Optimize before upload, save ~40%

2. **Search indexing** (compute): $1,200/month
   - Opportunity: Batch updates, save ~30%

3. **Payment retries** (Stripe API): $340/month
   - Opportunity: Better retry logic, save ~50%

### Cost Anomalies
⚠️ Database reads up 45% but traffic up only 12%
   Likely cause: Missing index or N+1 query

/gywd:impact --incidents

Show incident history and patterns.

## Incident Analysis

### Last 6 Months
| Severity | Count | MTTR | Trend |
|----------|-------|------|-------|
| P1 (Critical) | 2 | 52 min | → |
| P2 (Major) | 7 | 23 min | ↓ |
| P3 (Minor) | 15 | 45 min | → |

### By Code Area
| Area | Incidents | Last | Risk |
|------|-----------|------|------|
| src/api/checkout/ | 4 | 2 weeks | 🔴 High |
| src/services/payment/ | 3 | 1 month | 🟠 Medium |
| src/api/orders/ | 2 | 3 months | 🟡 Low |

### Root Cause Patterns
| Pattern | Count | Example |
|---------|-------|---------|
| Race condition | 3 | Concurrent cart updates |
| External service timeout | 3 | Stripe API latency |
| Database deadlock | 2 | Order + inventory lock |
| Memory exhaustion | 1 | Large export without streaming |

### Recommendations
1. **Add timeout handling** to all external calls
2. **Implement optimistic locking** for cart operations
3. **Add circuit breakers** to payment service

</commands> <integration> ## How Impact Integrates

During Planning

Planning Phase 4: Checkout Optimization

Impact context loaded:
- 47K requests/day (high impact changes)
- 2 P1 incidents in area (proceed carefully)
- $340/month in compute (optimization opportunity)

Recommendations:
- Add comprehensive tests before changes
- Implement feature flag for gradual rollout
- Have rollback plan ready

During Review

/gywd:challenge src/api/checkout.ts

Additional context from Impact:
- This file is in critical revenue path
- Last change caused P2 incident
- Suggest extra scrutiny on error paths

During Decisions

Decision: Add caching to product API

Impact analysis:
- Current: 892K requests/day at $0.0001/request
- With cache: ~70% cache hit expected
- Savings: ~$200/month
- Risk: Stale data, complexity

Recommendation: Proceed with 1-hour TTL

</integration> <configuration> ## Setup

Configure data sources in .planning/config.json:

{
  "impact": {
    "metrics": {
      "provider": "datadog",
      "api_key_env": "DD_API_KEY",
      "service_tag": "myapp"
    },
    "costs": {
      "provider": "aws",
      "profile": "production"
    },
    "incidents": {
      "provider": "pagerduty",
      "api_key_env": "PD_API_KEY"
    },
    "analytics": {
      "provider": "amplitude",
      "api_key_env": "AMPLITUDE_KEY"
    }
  }
}

If no integrations configured, falls back to:

• Git history analysis
• Manual annotations
• Estimated impact from code analysis </configuration>

<success_criteria>

• Retrieves production metrics per endpoint/file
• Shows business impact attribution
• Displays cost breakdown by code area
• Surfaces incident history
• Identifies patterns and anomalies
• Integrates with planning and review
• Falls back gracefully without integrations </success_criteria>