DevOps & Delivery Skills

Software delivery pipeline, read-only cloud monitoring, gated deployments, app provisioning and site reliability

5 skills

• App Provisioner

  AWS application provisioning from a git repo URL, gated end to end. Read/plan (no AWS, no spend): analyze_repo (detect stack/ports/env/db), plan_deployment (EC2+Docker+nginx+SSL+GitHub Actions golden-path plan + monthly cost estimate), create_stack, list_stacks, stack_status. Gated AWS actions provision_stack/deploy_release/teardown_stack require a human-approved stack (business manager approves in the console — the agent cannot) PLUS the operator flag (enabled by the operator); every AWS call is audited. Named operations only

  View skill →

• AWS Deploy

  Gated AWS deployments via named operations (no raw CLI): ECS force-redeploy, S3 site sync, CloudFront invalidation — requires the operator to enable deployments, plus per-call confirmation after Agent OS approval; full audit trail

  View skill →

• AWS Monitor

  Read-only AWS observability via named operations (no raw CLI): ECS service status, CloudWatch alarms and metrics, log tailing, EC2/RDS inventory, S3 buckets — every call audited

  View skill →

• Project Pipeline

  Software delivery pipeline (lead → requirements → offer → tech design → configuration → development → qa → uat → deployed → support) with enforced stage gates: approved offer before sending, approved technical document before development, qa/uat before deployment

  View skill →

• Site Ops Sentinel

  Storefront reliability: summarize uptime/checkout probes, evaluate Core Web Vitals, watch TLS expiry, and gate deployments. Production deploys are always a human release; staging green + rollback are preconditions.

  View skill →