aws-samples/sample-oh-my-aidlcops
5 skills5 installs80 starsGitHub
Install
npx skills add https://github.com/aws-samples/sample-oh-my-aidlcopsSkills in this repo
1ContainerizationContainerizes legacy applications with production defaults: a multi-stage Dockerfile, multi-arch (amd64/arm64) buildx builds, Trivy and grype security scanning, health checks, a non-root user, and minimal base images. It automates ECR push and generates ECS task definition or EKS Deployment manifest templates. A developer uses it after a to-be architecture is approved to package and prepare an app for AWS deployment. The docs are primarily in Korean.1installs2Cost GovernanceAggregates per-agent AWS cost using the AWS Pricing and Cost Explorer MCP servers, tracks monthly burn rate, and emits budget alerts at 80 and 95 percent thresholds. When usage patterns justify it, it recommends downgrading models from Opus to Sonnet to Haiku, and it vetoes deploys projected to exceed the monthly budget ceiling as a pre-flight gate for autopilot-deploy. It runs as a daily cron and reads budgets from a budget.yaml. The docs are primarily in Korean.1installs3Incident ResponseAutomates incident response from CloudWatch and Prometheus alarms. It classifies severity (SEV1 to SEV4), looks up runbooks, generates hypotheses paired with diagnostic MCP queries, and prepares remediation. SEV1 pages on-call immediately while the agent only diagnoses; SEV2 and SEV3 get a drafted remediation script that runs only after human approval. It records timelines, hypotheses and post-mortem drafts per incident. The docs are primarily in Korean.1installs4Self Improving Loopself-improving-loop watches an agent or skill's production traces in Langfuse and automatically analyzes quality regressions, cost anomalies, and failure patterns. It then generates prompt and skill fixes as PR drafts for human review. A developer runs it during periodic quality reviews or when a regression is detected, keeping the change gated behind human approval before deploy.1installs5Vllm Serving Setupvllm-serving-setup designs, deploys, and tunes vLLM inference serving on Amazon EKS for LLM workloads. A developer uses it to size GPU memory, pick a parallelism strategy, write Helm values, and validate a running model endpoint. It targets production agentic serving with autoscaling and Prometheus metrics.1installs