ams-senior-devops
暂无简介
- License 许可
- MIT
- 最近更新
- 2026-11-07
- Stars
- ★ 0
- Forks
- 1
03 SKILL.md 正文
Senior DevOps Engineer
Overview
Senior DevOps is applied reliability: making shipping fast and safe, and making systems explain themselves when they break. This skill captures that practice — pipeline design, immutable infrastructure, observability that actually helps at 3am, and incident response without heroics.
The through-line: automate the toil, measure what matters, and design for the failure you haven't had yet.
When to use
- Designing or fixing CI/CD pipelines.
- Choosing infrastructure approaches (containers, IaC, managed services).
- Setting up monitoring, alerting, and on-call practices.
- Responding to or learning from production incidents.
- Reducing deploy friction, lead time, or change-failure rate.
Core concepts
- Everything as code. Infrastructure, pipeline definitions, dashboards, and alerts live in version control and go through review like application code. Click-ops is unreviewable, unrepeatable, and unauditable.
- Immutable artifacts. Build once, promote the same artifact through environments. Rebuilding per environment means you're testing something different from what you ship.
- Progressive delivery. Small, frequent deploys behind feature flags, canary analysis, and automated rollback. The safest deploy is the one you do ten times a day — practice makes it boring.
- The four golden signals. Latency, traffic, errors, saturation — for every service. Alerts fire on symptoms (user-facing pain) with runbook links, not on every cause (a CPU spike nobody feels is not a 3am page).
- SLOs and error budgets. Define what "reliable enough" means numerically, then let the budget arbitrate the ship-vs-harden debate. No budget left? You fix reliability. Budget to spare? Ship.
- Blameless postmortems. Incidents are system failures, not people failures. The output is action items with owners and deadlines — "be more careful" is not an action item.
Practical workflow
- Map the path to production. Commit → build → test → artifact → stage → prod. Every manual step is a candidate for automation; every slow step gets measured.
- Codify infrastructure. Start with the network and compute baseline in Terraform/Pulumi/ CloudFormation; application config via environment, never baked into images; secrets via a secret manager, never in repos or env files committed anywhere.
- Build the deployment pipeline: lint + unit → build immutable image (pinned digests) → integration tests → deploy to staging (production-like data shape) → smoke tests → progressive prod rollout (canary 1% → 25% → 100% with automatic rollback on SLO breach).
- Instrument before you need it. Structured logs with correlation IDs, RED/USE metrics per service, distributed traces across boundaries, and dashboards for each service's golden signals.
- Define alerting tiers. Page (user impact now, actionable, runbook attached) vs ticket (degraded, investigate soon) vs log (informational). Every page must be actionable — if the response is "wait and see," it's not a page.
- Practice incidents. Game days / fire drills for the scary scenarios; postmortems within 48h of any significant incident; track action items to completion like product work.
Pipeline stage checklist:
[ ] Build is hermetic and reproducible (pinned base images, locked deps)
[ ] Same artifact promoted across environments (no rebuilds)
[ ] Secrets injected at runtime from secret manager
[ ] Rollback is one command / automatic on health-check failure
[ ] Deploy is progressive (canary or blue/green), not big-bang
[ ] Post-deploy smoke tests verify user-critical paths
[ ] Dashboard + alerts exist before first prod traffic
Common pitfalls
- Snowflake environments. "Works in staging" means nothing if staging differs from prod. Production-like data volume, config parity, and the same artifact — or your tests are theater.
- Alert fatigue. Paging on CPU thresholds and disk warnings trains on-call to ignore pages. Alert on user-impacting symptoms; tune ruthlessly; every false page is a bug in the alert.
- Manual deploys with extra steps. A wiki page titled "release process" with 30 manual steps is not a pipeline. If a human must remember the order, it will eventually go wrong at 2am.
- Secrets in repos. Even private repos leak (contractors, forks, history). Assume any committed secret is compromised: rotate and move to a manager.
- No rollback plan. "We'll roll forward" during an outage is optimism, not a plan. Fast, tested rollback beats heroic forward-fixing under pressure.
- Monitoring everything, observing nothing. 500 dashboards nobody opens. Start from questions ("is checkout healthy?") and build the minimum telemetry that answers them.
- Treating toil as inevitable. If on-call spends hours on repetitive tasks (log diving, manual scaling, cert renewals), that's engineering work waiting to be automated — budget for it.
04 同仓库更多 Skills
web-scrape
Turn web pages into LLM-ready markdown via the Firecrawl API (renders JS, anti-bot handling). Use when browser_open/fetch fails on JS-heavy pages, or when you need clean markdown/structured extraction from a site. Requires FIRECRAWL_API_KEY.
design-taste-frontend
tasteskill v2, the anti-slop frontend skill. Art direction for websites, distilled from a study of 247 curated gallery sites (One Page Love, Awwwards Sites of the Day). Deals researched directions (idea, composition, type, palette, photographs, section blueprints) from a library, enforces a strict no-slop house style with a real-browser inspector, and runs a look-and-fix loop on screenshots. Use this whenever you build, design, redesign, restyle, polish or review a landing page, marketing site, portfolio, product page, one-pager, hero or any public-facing web page, in any stack (HTML/CSS, React, Next.js, Vue, Svelte, Astro, Tailwind) — also when the user says a page looks generic, templated, "AI-made", sloppy, or asks to make it look designed, premium, award-level or less like every other site. Not for dashboards or app screens.
test-driven-development
Use when implementing any feature or bugfix, before writing implementation code
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes
requesting-code-review
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).