convez.ai
Agente
23seriy

devops-ai-workflows

23seriy/devops-ai-workflows

Ver no GitHub
BomPluginScore · de 0 a 100

Curated collection of AI-agent workflows, prompts & rules for DevOps/SRE — Kubernetes debugging, AWS audits, Terraform plan reviews, CI/CD triage, Dockerfile reviews, secrets scanning & incident response. Works with Windsurf, Cursor, Claude Code or any LLM.

Instalação

$npx convez-ai 23seriy/devops-ai-workflows

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
0
Acompanhando
0
Saúde
71%
Último commit
há 1 semana
Versão
v1.1.0
Issues abertas
0
Contribuidores
3

README

A tradução para o português deste README ainda está na fila.

devops-ai-workflows

A growing collection of AI-agent workflows, prompts, and rules for day-to-day DevOps / SRE / platform work.

Note: "workflows" here means Claude Code slash commands / AI-agent workflows — not GitHub Actions.

What's inside

FolderPurposeAudience
.claude/commands/Workflow definitions, auto-discovered as slash commands by Claude Code.Everyone
.claude/agents/Repo-maintenance subagents (e.g. workflow-author) invoked via the Agent tool.Maintainers
.claude/settings.jsonShared project settings: pre-approved read-only Bash(...) patterns + a deny list that blocks cluster/cloud mutations even if a workflow tries.Claude Code users
prompts/Reusable system / task prompts (incident triage, code review, post-mortem, etc.)Any LLM
rules/Reusable safety rule sets to load into Claude Code (via CLAUDE.md @-reference) or any other agentAny agent
scripts/Standalone shell scripts referenced by workflowsAnyone with a shell

Available workflows

Kubernetes

WorkflowSlash commandDescriptionPrerequisites
k8s-debug/k8s-debugGeneral-purpose, read-only cluster diagnostics across nodes, pods, workloads, networking, storage, RBAC, events, and resource pressure.kubectl. Optional: jq, metrics-server.
k8s-workload-debug/k8s-workload-debugDeep-dive on a single Deployment / StatefulSet / DaemonSet / Job / Pod: rollout, spec, probes, resources, logs, networking, storage, config.kubectl. Optional: jq, metrics-server.
k8s-rbac-audit/k8s-rbac-auditRBAC risk audit — wildcards, cluster-admin bindings, risky verb/resource combos, over-privileged ServiceAccounts, anonymous access.kubectl, jq. Optional: kubectl-who-can.
k8s-cost-hotspots/k8s-cost-hotspotsFind waste: over-provisioned workloads, missing requests/limits, idle workloads, orphan PVCs/PVs, idle LoadBalancers.kubectl, jq, metrics-server.
k8s-upgrade-readiness/k8s-upgrade-readinessPre-flight before a control-plane / node upgrade: deprecated APIs, version skew, PDB gaps, expiring certs, broken webhooks.kubectl. Optional: kubent or pluto, helm.
helm-release-debug/helm-release-debugDiagnose a stuck or failed Helm release: history, values diff, hook failures, rendered manifest vs cluster, workload health.helm v3, kubectl. Optional: jq, yq.
helm-chart-review/helm-chart-reviewReview a Helm chart for security, reliability, and best practices: resource specs, probes, security context, PDBs, anti-affinity, RBAC.Helm chart source. Optional: helm CLI.

AWS / Cloud

WorkflowSlash commandDescriptionPrerequisites
aws-account-audit/aws-account-auditRead-only AWS account security & hygiene audit: IAM, S3, EC2, RDS, CloudTrail, encryption, GuardDuty, SecurityHub.aws CLI. Optional: jq.
aws-cost-quickscan/aws-cost-quickscanFind AWS cost waste: idle EC2/RDS, unattached EBS, old snapshots, expensive log groups, NAT data processing, missing Savings Plans.aws CLI, Cost Explorer enabled. Optional: jq.
aws-vpc-debug/aws-vpc-debugDiagnose VPC connectivity: trace path across SGs, NACLs, route tables, NAT/IGW/TGW, VPC endpoints, DNS, and flow logs.aws CLI. Optional: jq, dig.
aws-iam-policy-review/aws-iam-policy-reviewExplain an IAM policy and flag risks: admin-equivalent access, privilege escalation paths, wildcard actions, missing conditions.aws CLI. Optional: jq.

IaC

WorkflowSlash commandDescriptionPrerequisites
terraform-plan-review/terraform-plan-reviewExplain a Terraform plan and flag risky changes: destroys, replacements, security group mutations, IAM changes, blast radius.terraform plan output. Optional: terraform CLI, jq.

Containers & CI/CD

WorkflowSlash commandDescriptionPrerequisites
ci-debug/ci-debugDiagnose a failing CI/CD pipeline: parse build logs from Jenkins, GitHub Actions, GitLab CI, or Bitbucket Pipelines. Root cause analysis and fix suggestions.Build log output. Optional: repo source, CI config file.
jenkins-pipeline-review/jenkins-pipeline-reviewReview Jenkinsfile / shared-library Groovy for security risks, anti-patterns, missing error handling, credential leaks, CPS issues, and build config cross-references.Jenkinsfile(s) or vars/*.groovy. Optional: repositories_v2.json.
release-checklist/release-checklistPre-release safety gate: scope, deploy order, rollback, tests, monitoring, and communication before production release.PR/diff summary. Optional: test results, plans, diffs.
dockerfile-review/dockerfile-reviewReview Dockerfiles for security, size, caching, and best practices. Flags CVE-prone bases, leaked secrets, missing health checks.Dockerfile(s). Optional: docker, trivy.

Security

WorkflowSlash commandDescriptionPrerequisites
secrets-leak-scan/secrets-leak-scanScan git repo history for leaked secrets: API keys, passwords, tokens, private keys. Uses gitleaks, trufflehog, or regex fallback.Git repo. Optional: gitleaks, trufflehog.
repo-health/repo-healthAudit repository hygiene: README, license, CI, branch/release hygiene, tracked secrets, ownership, and automation gaps.Local git repo. Optional: gh, jq.

Observability & Incident

WorkflowSlash commandDescriptionPrerequisites
incident-triage/incident-triageGuided first 15 minutes of a production incident: timeline, blast radius, evidence gathering, mitigation suggestions.Access to affected environment.

More on the way — see Roadmap.

Prompts

Reusable system prompts you can paste into any AI agent for common DevOps tasks:

PromptWhat it does
incident-commanderPuts the AI in incident-commander mode: timeline, blast radius, action tracking, status updates.
postmortem-writerGenerates a blameless post-mortem from incident notes: timeline, root cause, impact, action items.
code-review-devopsReviews IaC / pipeline / Docker / K8s code with a security-first DevOps lens.
pr-descriptionGenerates a PR description from a diff: what, why, how, testing, risk, rollback plan.
explain-like-a-seniorExplains infrastructure code to junior engineers: what it does, why, gotchas, and how it fits together.
runbook-from-incidentConverts incident notes or post-mortems into reusable runbooks with diagnosis, mitigation, escalation, and follow-up steps.

Rules

Reusable, agent-agnostic safety rule sets. Reference them from a project's CLAUDE.md (e.g. @rules/kubernetes.md), paste into a system prompt, or include as context:

Rule fileWhat it does
devops-agent.mdSafety guardrails for AI in DevOps repos: never modify prod without confirmation, prefer read-only, never hardcode secrets, always check context, GitOps awareness, multi-repo coordination.
terraform.mdTerraform-specific: state safety, ForceNew attribute warnings, provider/module pinning, workspace safety, import workflow, prevent_destroy reminders.
kubernetes.mdKubernetes-specific: context verification, dry-run first, Helm safety, ArgoCD/GitOps awareness, secret handling, debugging approach, RBAC best practices.

Scripts

Standalone shell utilities referenced by workflows or useful on their own:

ScriptUsage
k8s-snapshot.sh./k8s-snapshot.sh [namespace|all] [output-dir] — dump cluster state (nodes, pods, events, services, top) to a timestamped Markdown file.
aws-whoami.sh./aws-whoami.sh [profile] — quick AWS identity check: caller, region, account alias, org, SSO role.
stale-branches.sh./stale-branches.sh [days] [--remote] — list git branches older than N days with last commit info.
validate-repo.sh./scripts/validate-repo.sh — validate workflow frontmatter, README links, script executability, and optional lint checks.

Using a workflow

In Claude Code

Clone the repo and run Claude Code from the repo root. Every workflow under .claude/commands/ is auto-discovered as a slash command — .claude/commands/k8s-debug.md is invoked as /k8s-debug, etc. The argument-hint in each file's frontmatter is shown inline as you type.

CLAUDE.md @-imports the rule files under rules/, so the safety guardrails for Kubernetes, Terraform, and general DevOps work load automatically whenever you run Claude Code in this directory.

.claude/settings.json pre-approves read-only bash patterns commonly used while developing here (validate-repo, find, grep, git status/diff/log) and denies cluster/cloud mutations (kubectl apply|delete|scale, helm install|upgrade, terraform apply|destroy, aws iam create|delete|put, etc.) — keeping the repo's read-only contract enforced at the tool level.

In other AI agents

Open the matching file in .claude/commands/ and either:

  • paste the relevant section into the agent's chat, or
  • include the file as context and ask the agent to follow it.

As a plain human workflow

Every workflow is just Markdown with shell commands. You can run the steps yourself in a terminal — no AI required.

Slash command frontmatter

Every workflow declares two YAML keys:

---
description: One-sentence summary shown in the slash-command picker.
argument-hint: "REQUIRED=<value> [OPTIONAL=default]"
---

description shows up in the /-picker. argument-hint mirrors the Inputs section so users see expected args inline. See CONTRIBUTING.md for examples.

Repo layout

devops-ai-workflows/
├── .claude/
│   ├── commands/            # Workflow definitions (Claude Code slash commands)
│   ├── agents/              # Repo-maintenance subagents (workflow-author, ...)
│   └── settings.json        # Shared allow/deny permissions
├── prompts/                 # Reusable LLM prompts
├── rules/                   # Editor/agent rule files (auto-loaded via CLAUDE.md)
├── scripts/                 # Standalone shell helpers
├── CLAUDE.md                # Claude Code project instructions
├── CONTRIBUTING.md
├── LICENSE
└── README.md

Roadmap

Ideas I plan to add (PRs welcome):

AWS / cloud

  • /aws-eks-debug — bridge EKS + Kubernetes: node groups, OIDC, add-ons, IAM roles for service accounts
  • /aws-rds-health — RDS/Aurora diagnostics: events, metrics, parameter groups, replication lag
  • /aws-lambda-debug — Lambda diagnostics: errors, throttles, DLQ, VPC/ENI, CloudWatch logs
  • /aws-ecs-service-debug — ECS/Fargate service rollout failures: task events, target group health, IAM roles

IaC

  • /terraform-state-debug — diagnose locks, drift, orphans
  • /iac-secrets-scan — repo-wide hardcoded-secret sweep

Containers & CI/CD

  • /image-cve-triage — prioritise CVE scanner output by exploitability + fix availability
  • /github-actions-review — security review of GitHub Actions workflow files

Observability & incident

  • /prometheus-query-helper — intent → PromQL with rationale
  • /log-pattern-extract — cluster repeated errors out of a log dump
  • /postmortem — blameless post-mortem from a transcript
  • /runbook-from-incident — turn a resolved incident into a reusable runbook

Networking / database

  • /dns-debug — multi-resolver dig, propagation, DNSSEC
  • /tls-cert-audit — chain inspection, expiry, weak ciphers across a list of hosts
  • /postgres-health — bloat, long queries, replication lag, missing indexes
  • /redis-health — memory pressure, slow log, persistence config, eviction patterns
  • /db-migration-review — flag risky migration patterns

Security & repo hygiene

  • /cve-impact-assessment — given a CVE, check whether your stack is affected
  • /repo-health — README, license, CI, branch protection, stale branches
  • /dependency-upgrade-plan — group outdated deps by risk and suggest batching

Contributing

See CONTRIBUTING.md. The short version:

  1. Add the canonical workflow to .claude/commands/<name>.md.
  2. Update the Available workflows table in this README.
  3. Keep workflows read-only by default. Anything mutating must be opt-in (e.g. a DEEP=yes flag) and clearly flagged.

License

MIT — use freely, attribution appreciated but not required.

Ver o original no GitHub

Bom para

Relacionados