Skip to main content
The Runlayer Infrastructure Upgrader helps your local AI agent prepare a new deployment or upgrade an existing installation. It reads current public deployment and release docs, asks about your environment, and produces configuration for review. This skill covers AWS ECS and EKS. GKE/GCP deployments require a separate workflow; see Kubernetes deployment for public guidance. Download v0.3.0 · SHA-256 checksum

​
Install

Extract the ZIP and keep the runlayer-infra-upgrader folder together. Extraction alone does not register the skill with your agent. Choose one installation method: Check for an existing installation before copying; replace it deliberately rather than leaving multiple versions. Claude Code’s skill locations are separate from this repository’s .agents source and module-bundle paths. Do not assume .agents/runlayer-infra-upgrader/ will be discovered automatically. For an agent with local file access, this explicit-path prompt works without automatic skill discovery (replace /path/to with the actual extracted location):
For an upgrade, replace the request with your deployment type and current/target versions. For EKS, the skill uses the Runlayer Operator and RunlayerInstance. It identifies legacy installations before proposing a migration. New ECS Terraform module archives also include this skill under .agents/skills/runlayer-infra-upgrader/. Older archives may contain the previous skill under .claude/skills/; use this standalone download for the revised workflow. If you installed the previous skill globally, replace that installation to avoid loading both copies. Installation paths vary by agent. The skill needs no Runlayer MCP connection or access to Runlayer’s source repository. Your agent needs HTTPS access to public docs and local file access to generate files. Validation also needs the selected platform’s tools, artifact access and your cloud identity. Your agent’s data handling and permission settings still apply.

​
Prepare

Have your deployment type, account/region, network and DNS ownership, state/backend conventions, availability requirements and existing repository ready. For upgrades, also identify installed versions and customizations. You can start with missing information; the agent will label unresolved values and access requirements. Get onboarding inputs through Runlayer-provided inputs. Deliver secret values through your secret manager or environment, never the chat. This includes the Distribution reader credential and its registration requirements.

​
What the skill does

  • Discovers guidance through llms.txt.
  • Recommends latest published releases for new setups, respects explicit pins and checks compatibility against the selected module or chart.
  • Prepares ECS Terraform or EKS infrastructure/operator configuration using the current public examples, rather than a bundled snapshot of deployment defaults.
  • Maps upstream changes to customized installations for upgrades.
  • Reports generated files, source URLs, chosen versions, validation results and unresolved prerequisites in a deployment handoff.
The download includes workflow instructions and a license, not a copy of the deployment manuals. If current documentation cannot be retrieved, the agent can prepare a draft but must identify what it could not verify. Applying infrastructure, changing state or migrating controllers requires explicit authorization. Formatting, rendering or a successful plan does not establish runtime health. Follow ECS deployment, Runlayer Operator and Updates for deployment and verification.