AgentSkillsCN

runbook

生成服务、程序或事故响应的操作手册

SKILL.md
--- frontmatter
name: runbook
description: Generate operational runbooks for services, procedures, or incident response

Operational Runbook Generator

Current Time: !date K8s Context: !kubectl config current-context

Generate operational runbooks for services, procedures, or incident response. Investigates the codebase and infrastructure to produce accurate, actionable procedures. Documents to Obsidian.

Input

  • •Topic: service name, operation type, or incident scenario
  • •Scope: deployment, scaling, failover, maintenance, troubleshooting
  • •Optional: specific scenarios to cover

Investigation Strategy

Launch parallel investigation tracks:

Track 1: Codebase Exploration (explore agent)

  • •Identify service entry points and configuration
  • •Find health check endpoints
  • •Map dependencies (databases, caches, external services)
  • •Locate logging and metrics instrumentation
  • •Find existing scripts or automation

Track 2: Infrastructure Analysis (inferred agent: k8s/sre/gcp-dev)

  • •Review deployment manifests
  • •Identify scaling configuration
  • •Map service dependencies
  • •Find monitoring and alerting setup
  • •Review backup and recovery procedures

Track 3: External Research (librarian agent)

  • •Find operational best practices for the service type
  • •Research common failure modes
  • •Identify industry-standard procedures

Output

Write to Obsidian via obsidian_append_content at: $OBSIDIAN_PATH/Runbooks/YYYY-MM-DD-service-operation.md

Note: $OBSIDIAN_PATH must be a vault-relative path (e.g., Projects/myapp), set per-project via direnv. The obsidian_append_content tool expects paths relative to the vault root.

Document Structure

Use this template for the Obsidian document:

@~/.config/opencode/templates/runbook.md

Behavior

  1. •Parse topic to identify service and operation scope
  2. •Infer appropriate infrastructure agent (k8s, sre, gcp-dev)
  3. •Launch explore, librarian, and inferred agent in parallel
  4. •Extract configuration, endpoints, and dependencies from codebase
  5. •Identify common operations and failure modes
  6. •Generate step-by-step procedures with actual commands
  7. •Write runbook to Obsidian via obsidian_append_content with auto-generated filename: YYYY-MM-DD-service-operation.md

$ARGUMENTS