/prd-v08-runbook-creation

name: prd-v08-runbook-creation
description: >
  Create operational playbooks for incident response, deployments, and maintenance during PRD v0.8 Deployment & Ops.
  Triggers on requests to create runbooks, document procedures, or when user asks "how do we handle incidents?",
  "runbook", "operational procedures", "on-call guide", "incident response", "maintenance procedures".
  Outputs RUN- entries with step-by-step operational procedures.
context: fork
allowed-tools:
  - Read
  - Write
  - Edit
  - Glob
  - Grep
  - Bash

execution_modes:
  default: standard
  supports: [quick, standard, deep]

Runbook Creation

Position in workflow: v0.8 Release Planning → **v0.8 Runbook Creation** → v0.8 Monitoring Setup

Execution Mode

Default is **standard**. See [`.claude/rules/08-skill-execution-modes.md`](../../rules/08-skill-execution-modes.md) for selection logic.

| Mode | What this skill produces | |------|--------------------------| | **quick** | 2–3 runbooks covering the top RISK-/MON- pairs | | **standard** | Incident + deployment + maintenance runbooks linked to MON- and DEP-, with escalation paths | | **deep** | Full matrix + drill schedule + recovery scenarios + tabletop exercise outputs |

Consumes

This skill requires prior work from v0.8 Release Planning and earlier stages:

• **DEP-\* deployment entries** (from v0.8 Release Planning) — Deployment procedures from DEP- rollback/validation sections inform RUN- deployment and recovery runbooks
• **RISK-\* risk entries** (from v0.5 Risk Discovery) — High/medium RISK- entries must have response runbooks; mitigations become procedures
• **MON-\* monitoring specifications** (planned from v0.8 Monitoring Setup, or anticipated) — Key alerts from MON- (before formalization) inform incident response runbooks; runbooks will be referenced FROM monitoring
• **ARC-\* architecture decisions** (from v0.6 Architecture Design) — System structure (single service vs microservices, databases, integrations) determines incident scope and escalation paths
• **TECH-\* technology stack** (from v0.5 Technical Stack Selection) — Technology choices (database, cloud provider, APM tools) determine specific commands and tools in runbook procedures
• **API-\* endpoint contracts** (from v0.6 Technical Specification) — For reference if incident involves specific endpoints or payloads

This skill assumes DEP- entries are complete with rollback procedures and post-deploy validation steps defined.

Produces

This skill creates/updates:

• **RUN-\* entries** (operational runbooks, category-based) — Step-by-step procedures for incident response, deployment execution, maintenance tasks, recovery from failures, and escalation paths
• **Incident response matrix** — Mapping of scenarios (connection pool exhaustion, latency spike, deployment failure, etc.) to RUN- procedures
• **Runbook cross-reference** — Links from RUN- procedures to DEP- rollback conditions, anticipated MON- alerts, and RISK- entries they address

All RUN- entries are **operational procedures**, not confidence-based. They are:

• **Executable** (numbered steps with specific commands and tools)
• **Verifiable** (each step has a verification check confirming success)
• **Scoped** (explicit "Handles" and "Does NOT handle" sections)
• **Escalatable** (clear escalation paths and contact information)
• **Tested** (should be drilled regularly; includes "Last Tested" date)

Example RUN- entries:

RUN-001: Database Connection Pool Exhaustion
Category: Incident
Trigger: MON-005 alert (connection pool >90%) — from v0.8 Monitoring Setup
Owner: Backend Team
Last Tested: 2025-02-20

## Scope
- **Handles**: Connection pool saturation, slow queries causing pooling
- **Does NOT handle**: Database server crash (see RUN-010), Network failure (see RUN-011)

## Prerequisites
- [ ] Access to AWS RDS console
- [ ] PostgreSQL read credentials in LastPass
- [ ] PagerDuty access for escalation
- [ ] Datadog dashboard access (MON-005 source)

## Procedure

### Step 1: Verify Alert
Check current connection pool status:

Commands:
\`\`\`sql
SELECT count(*) FROM pg_stat_activity WHERE state = 'active';
SELECT * FROM pg_stat_activity WHERE state = 'active' ORDER BY query_start;
\`\`\`

Verification:
- [ ] Connection count ≥90% of max pool (check DEP-001 pool config)

### Step 2: Identify Problematic Queries
Find long-running or blocked queries:

Commands:
\`\`\`sql
SELECT pid, now() - pg_stat_activity.query_start AS duration, query
FROM pg_stat_activity
WHERE (now() - pg_stat_activity.query_start) > interval '5 minutes';
\`\`\`

Verification:
- [ ] At least one query identified running >5 minutes

### Step 3: Kill Problematic Queries (if safe)
Only kill queries that are clearly stuck:

Commands:
\`\`\`sql
SELECT pg_terminate_backend(pid) FROM pg_stat_activity
WHERE pid = <problematic_pid>;
\`\`\`

Verification:
- [ ] Connection count dropping within 2 minutes
- [ ] MON-005 alert resolving

### Step 4: Investigate Root Cause
- Check recent deployments (last 24h via git log)
- Review application logs for query patterns (Datadog logs)
- Check for missing indexes on recent queries

## Escalation
- **When to escalate**: Issue persists >15 minutes, data integrity concern, cannot kill queries safely
- **Who to contact**: Database Team Lead (Slack: @db-team, PagerDuty)
- **What to provide**: Timeline, queries identified, actions taken, connection count trend

## Post-Incident
- [ ] Document incident timeline (who was paged, when actions taken)
- [ ] File ticket for query optimization if needed
- [ ] Update this runbook if steps were wrong/missing
- [ ] Schedule team drill of this runbook within 1 week if escalated

Linked IDs: MON-005 (alert), DEP-001 (pool config), RISK-008 (data integrity)

---

RUN-002: Production Deployment Procedure
Category: Deployment
Trigger: Scheduled release when all DEP- criteria met
Owner: DevOps Team
Last Tested: 2025-02-18

## Scope
- **Handles**: Standard production deployments (mainlin