/devops
DevOpsModel-invoked

When deploy or CI breaks: search Knowledge with `intent: incident`, confirm cause, fix, verify, then close with SEV or post-mortem templates when severity warrants it.

When to use

Vercel or build failures, monorepo deploy traps, CI infra changes — search Knowledge (`intent: incident`) before changing config.

Prerequisites

  • A deploy/CI/infra symptom (logs, failing job, host error).
  • Knowledge MCP or kit site Knowledge available when searching incidents.

How to use

  1. Invoke /devops with the symptom (e.g. Vercel deploy failed on TanStack Start).
  2. Require Knowledge search with `intent: incident` before changing config.
  3. Confirm cause, apply the fix, and verify green deploy/CI.
  4. For SEV1/SEV2, close with SEV or post-mortem templates (`techlead` owns SEV).

Done when

Deploy/CI is verified green (or rolled back safely) and severity paperwork is done when required.

Pipeline

Upstream
App deploy docs (e.g. `apps/*/DEPLOY.md`)
Downstream
Verified green deploy/CI (+ post-mortem when SEV1/SEV2)

Boundaries

Not application feature code — operational playbooks and infra config. Design seams → `arch`; feature work → /dev.

Sample prompt

prompt
/devops

Vercel deploy failed on TanStack Start monorepo

Agent

PRINCIPAL ENGINEER · TECH LEAD

Owns: incident knowledge, stack profiles, deploy/CI fixes, SEV/post-mortem

Use the developer to [symptom] — or techlead for SEV ownership
developertechlead
← Skill catalog