production-readiness（生产准备）

Production Readiness (Meta-Skill)

Coordinates all operational concerns into a single readiness review. Instead of duplicating domain expertise, this skill routes to specialized skills and agents for each area, then synthesizes results into a unified go/no-go assessment.

Installation

OpenClaw / Moltbot / Clawbot

npx clawhub@latest install production-readiness

Purpose

Ensure a service is production-ready by systematically checking every operational concern — logging, error handling, performance, security, deployment, testing, and documentation — before traffic hits it.

A production-ready service:

• Fails gracefully under load and partial outages
• Observes itself with structured logs, metrics, and traces
• Recovers automatically from transient failures
• Communicates health to orchestrators and operators
• Documents operations so on-call engineers can respond without tribal knowledge

When to Use

Orchestration Flow

Run each area sequentially or in parallel. Each step delegates to a specialized skill or agent — this skill does not re-implement their logic.

┌─────────────────────────────────────────────────┐
│              Production Readiness Review         │
├─────────────────────────────────────────────────┤
│                                                  │
│  1. Logging & Observability ──► logging-observability skill
│  2. Error Handling ───────────► error-handling-patterns skill
│  3. Performance ──────────────► performance-agent
│  4. Security ─────────────────► security-review meta-skill
│  5. Deployment ───────────────► deployment-agent + docker-expert skill
│  6. Testing ──────────────────► testing-workflow meta-skill
│  7. Documentation ────────────► /generate-docs command
│                                                  │
│  ──► Synthesize results into go/no-go report     │
└─────────────────────────────────────────────────┘

Step Details

1. Logging & Observability — Structured logging, log levels, correlation IDs, metrics endpoints, distributed tracing, alerting rules
2. Error Handling — Global error boundaries, retry policies, dead-letter queues, error classification, user-facing error messages
3. Performance — Load testing results, P95/P99 latency baselines, memory/CPU profiling, database query analysis, caching strategy
4. Security — Auth/authz verification, input validation, dependency audit, secrets management, OWASP top-10 review
5. Deployment — Container hardening, rollback strategy, blue-green/canary configuration, infrastructure-as-code review
6. Testing — Unit/integration/e2e coverage, contract tests, chaos/failure injection, smoke tests in staging
7. Documentation — API docs, runbooks, architecture diagrams, on-call playbooks, ADRs for key decisions

Skill Routing Table

Routing rule: Read the target skill first, follow its instructions, then return results here for synthesis.

Production Readiness Checklist

Health & Lifecycle

• [ ] Health check endpoint (/healthz or /health) returns dependency status
• [ ] Readiness probe distinguishes "starting" from "ready to serve"
• [ ] Liveness probe detects deadlocks and unrecoverable states
• [ ] Graceful shutdown drains in-flight requests before exit
• [ ] Startup probe allows for slow initialization without false restarts

Resilience

• [ ] Circuit breakers on all external service calls
• [ ] Retry with exponential backoff and jitter on transient failures
• [ ] Rate limiting configured per endpoint and per client
• [ ] Backpressure mechanisms prevent cascade failures under load
• [ ] Timeouts set on every outbound call (HTTP, DB, queue)
• [ ] Bulkhead isolation separates critical from non-critical paths

Configuration & Secrets

• [ ] All configuration externalized (env vars, config service, or feature flags)
• [ ] No secrets in code, images, or environment variable defaults
• [ ] Secrets loaded from a vault (e.g., AWS Secrets Manager, HashiCorp Vault)
• [ ] Configuration changes do not require redeployment
• [ ] Feature flags in place for high-risk changes

Data Safety

• [ ] Backup strategy defined and tested (RPO/RTO documented)
• [ ] Restore procedure verified in a non-production environment
• [ ] Database migrations are backward-compatible and reversible
• [ ] Data retention policies implemented and enforced

Operational Readiness

• [ ] Runbooks exist for top 5 most likely failure scenarios
• [ ] SLOs defined (availability, latency, error rate) with error budgets
• [ ] SLAs communicated to dependent teams or customers
• [ ] On-call rotation staffed and escalation path documented
• [ ] Dashboards show golden signals (latency, traffic, errors, saturation)
• [ ] Alerting rules configured with appropriate thresholds and severity

Maturity Levels

Progression Guidance

• L1 → L2: Add structured logging, metrics endpoint, and a CI/CD pipeline. Write runbooks for known failure modes.
• L2 → L3: Instrument distributed tracing. Add circuit breakers to external calls. Define SLOs and set up on-call.
• L3 → L4: Implement canary deployments. Adopt error budgets. Run regular game days. Build predictive alerting.

Incident Response

On-Call Rotation

• Minimum two engineers per rotation (primary + secondary)
• Handoff includes review of recent deploys, open issues, and known risks
• Escalation targets defined: primary → secondary → engineering lead → VP Eng

Escalation Matrix

Postmortem Template

## Incident: [Title]
**Date:** YYYY-MM-DD | **Duration:** X hours | **Severity:** SEV-N

### Summary
One-paragraph description of what happened and impact.

### Timeline
- HH:MM — First alert fired
- HH:MM — Engineer paged, investigation started
- HH:MM — Root cause identified
- HH:MM — Mitigation applied
- HH:MM — Full resolution confirmed

### Root Cause
What broke and why. Link to code/config change if applicable.

### Impact
- Users affected: N
- Revenue impact: $X (if applicable)
- SLO budget consumed: X%

### Action Items
| Action | Owner | Due Date | Status |
|--------|-------|----------|--------|
| Fix X  | @eng  | YYYY-MM-DD | Open |

### Lessons Learned
- What went well
- What went poorly
- Where we got lucky

NEVER Do

1. NEVER skip health checks — every service must expose health endpoints; no exceptions for "simple" services
2. NEVER store secrets in code or container images — use a secrets manager; never default env vars with real values
3. NEVER deploy without a rollback plan — if you cannot roll back in under 5 minutes, you are not ready to deploy
4. NEVER ignore error budget violations — when the error budget is exhausted, freeze feature work and fix reliability
5. NEVER treat logging as optional — a service without structured logging is a service you cannot debug at 3 AM
6. NEVER go to production without runbooks — if on-call cannot resolve the top 5 failure modes without the original author, the service is not production-ready