Skill: Observability Pack Implementer

Purpose

Establish structured logging, tracing, and metrics conventions that enable operators to diagnose production issues without source code access.

When to Use This Skill

• •Setting up new services or applications
• •Adding observability to existing services
• •Implementing correlation ID propagation
• •Adding metrics or traces to critical paths
• •Enforcing redaction policies

Steps

1) Adopt structured logging schema

Ensure all logs are JSON with required fields:

• •timestamp, level, message, service, env, version, request_id
• •Recommended: component, op, duration_ms, error

Reference: observability/logging_schema.md

2) Implement correlation ID propagation

• •Generate request_id at entry (HTTP handler, job runner)
• •Pass request_id to all downstream operations
• •Ensure every log includes request_id

Example (JavaScript):

import { createLogger, ensureRequestId } from './obs.js';

app.use((req, res, next) => {
  const requestId = ensureRequestId(req.headers);
  const logger = log.child({ request_id: requestId });
  req.logger = logger;
  res.setHeader('X-Request-Id', requestId);
  next();
});

3) Add RED/USE metrics

• •Rate: requests_total (counter)
• •Errors: request_errors_total (counter)
• •Duration: request_duration_ms (histogram)

Example:

app.use((req, res, next) => {
  const start = Date.now();
  res.on('finish', () => {
    const duration = Date.now() - start;
    metrics.recordRequest(req.method, req.route, res.statusCode, duration);
  });
  next();
});

4) Implement redaction policy

Apply redaction to denylisted keys in logs:

• •Never emit: passwords, tokens, API keys, raw PII
• •Use allowlist patterns: internal IDs, hashes, truncated values

Reference: observability/redaction_policy.md

5) Add tracing (optional, for complex flows)

Use OpenTelemetry or similar:

• •Create spans for each operation (HTTP, DB, external API)
• •Include correlation IDs in spans
• •Never attach sensitive payloads

Reference: observability/tracing_conventions.md

Quality Checklist

• • All logs are JSON with required fields
• • request_id is generated and propagated end-to-end
• • RED metrics are recorded
• • Redaction is enforced for sensitive keys
• • Spans exist for external calls (if using traces)
• • Tests verify logs and metrics are emitted

Verification Commands

npm test 2>&1 | head -20 | grep -o '"service":"[^"]*"'
npm test 2>&1 | jq '.request_id' | sort | uniq
npm test 2>&1 | grep -i "password\|token" || echo "Clean"

KAIZA-AUDIT Compliance

When using this skill, your KAIZA-AUDIT block must include:

• •Scope: Services or modules with new observability
• •Verification: Sample log output, correlation ID propagation confirmed
• •Results: Redaction verified, no sensitive data leaked