Guide · Agent skill
How to use the Cloud Run Alert Configuration agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Cloud Run Alert Configuration skill teaches an AI agent to design monitoring alerts for Google Cloud Run, Google's service for running containers. Once installed, the agent writes Terraform files that create Cloud Monitoring alert policies using PromQL queries. This guide is for platform and SRE engineers, and developers, who want a reviewed starting set of alerts they can adjust. By the end you will know how to install the skill, ask for alerts in plain language, and check what the agent produces before you apply it.
Cloud Run Alert Configurationgoogle/skills · Apache-2.0 Download the skill Quick start
- 1Install the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. In the Claude apps, upload it under Settings, Capabilities, Skills.
- 2Confirm the tools. Make sure Terraform is installed. Install the gcloud CLI if you want the agent to discover your project and services automatically.
- 3Name your target. Ask for alerts for a specific Cloud Run service, job or worker pool. Naming it skips discovery and speeds up the response.
- 4Review the generated Terraform. Read the google_monitoring_alert_policy resources, the variables and the user_labels block. Check that project uses var.scoping_project_id.
- 5Check the defaults. Ask the agent to state the default thresholds and explain each PromQL query in plain English.
- 6Set notification channels. Provide your notification channel IDs through var.notification_channels so alerts reach the right people.
- 7Adjust and apply. Override any default through variables or follow-up prompts, validate against real traffic, then apply the configuration in a Google Cloud project with Cloud Monitoring access.
What is an agent skill and what does this one do?
An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches the skill's description. This skill's main file is SKILL.md, and it is supported by reference guides for services, jobs, worker pools and attribution.
When the skill is active, the agent covers error rates (4xx and 5xx), request latency, container instance saturation, container CPU and memory use, billable instance time, job execution status and worker pool queue backlog. It applies default thresholds based on SRE practice and explains in plain English what each alert means and how it can be tuned.
Install the skill from the ZIP or the Claude apps
Download the ZIP for this skill from this page. You have two options. For Claude Code, unzip it into ~/.claude/skills/ to make it available in all your projects, or into a project's .claude/skills/ folder to limit it to that project.
If you use the Claude apps, upload the skill under Settings, then Capabilities, then Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into .claude/skills/ inside your project
- Claude apps: Settings, Capabilities, Skills, then upload
Check the requirements before you start
The skill uses Terraform and the gcloud CLI. Terraform is needed to write and check the configuration. The gcloud CLI is optional and is only used to find your project and services when you do not name a target. To apply the generated configuration you need a Google Cloud project with Cloud Monitoring access.
The skill does not run scripts and does not use the network on its own. Check that Terraform is installed in the environment the agent works in.
- Terraform (required)
- gcloud CLI (optional, used for discovery when no target name is given)
- Google Cloud project with Cloud Monitoring access to apply the configuration
Trigger the skill with a natural request
Name the Cloud Run service, job or worker pool in your request. The skill then skips file and CLI discovery and goes straight to generating alerts, with the project ID and target name placed in Terraform variables. If you do not name a target, the agent first looks for Cloud Run resources in the .tf files in the current directory, then tries gcloud. If gcloud fails or Terraform is missing, it outputs parameterized Terraform using default variable values.
The skill is meant for Cloud Run. It says to use the gke-alert-configuration skill for GKE workloads and does not cover Compute Engine VMs.
- Create Cloud Run alerts for the web-frontend service in project my-gcp-project
- Write Terraform alert policies for the nightly-reconciliation Cloud Run job
- Set up worker pool alerts for order-processor with notification channels from variables
- Review my Cloud Run service .tf files and suggest baseline alerts
What the agent generates for each workload type
The agent picks a reference guide based on the workload. HTTP services get availability SLO alerts on 5xx errors, P95 and P99 latency, 4xx client errors, instance saturation, CPU and memory use, traffic drops and surges, and billable instance time. Jobs get an immediate alert on any failed execution. Worker pools get a standard suite of four policies: task success SLO fast and slow burn, backlog estimated time to drain, and message age SLA.
Every alert policy includes a user_labels block with created-with-google-skill set to cloud-run-alert-configuration, sets project to var.scoping_project_id so it can target one project or a metrics scope, and uses alert_strategy with auto_close set to 604800s. Notification channels come from var.notification_channels.
- HTTP services: 5xx SLO, P95/P99 latency, 4xx, saturation, CPU/memory, traffic anomalies, billable time
- Jobs: failed executions (result="failed") with duration 0s
- Worker pools: task success SLO fast and slow burn, backlog ETD, message age SLA
user_labels = {
created-with-google-skill = "cloud-run-alert-configuration"
}Review the defaults and PromQL before applying
The skill states the default thresholds it uses and says each can be overridden through Terraform variables or in your prompt. Ask the agent to list the defaults and explain the PromQL logic in plain English. Check these before you run terraform apply.
For short lookback windows of 25 hours or less, the skill sets duration to 300s to absorb brief spikes and scale-up lag, except job failure alerts, which use 0s. For lookbacks longer than 25 hours, such as 3- or 7-day slow burn alerts, duration is omitted because Cloud Monitoring rejects it there. The defaults are generic, so validate them against your own traffic before production use.
- Ask for the list of default constants and their variable names
- Confirm notification channels are set
- Test thresholds against real traffic before relying on them
What people use it for
- Adding a starting set of 5xx, latency and saturation alerts to a new Cloud Run web service
- Creating immediate failure alerts for a scheduled Cloud Run job
- Setting up task success and backlog alerts for a Cloud Run worker pool
- Turning an existing Terraform setup for Cloud Run monitoring into a reviewed baseline with user labels
Frequently asked questions
How do I install the cloud run alert configuration skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. In the Claude apps, upload it under Settings, Capabilities, Skills.
Is the cloud run alert configuration skill free?
The skill is published under the Apache-2.0 licence. The source does not mention pricing, so check the licence text for terms. Google Cloud and Cloud Monitoring usage is billed separately by Google.
What do I need to use this skill?
You need Terraform. The gcloud CLI is optional and only used for discovery when you do not name a target. To apply the output you need a Google Cloud project with Cloud Monitoring access.
Does this skill work for GKE or Compute Engine?
No. The skill says GKE workloads should use the gke-alert-configuration skill, and it does not cover Compute Engine VMs. Results for those workloads would be wrong.
Does the skill run scripts or use the network?
According to its listing, it runs no scripts and uses no network. It does tag gcloud commands run in the terminal with a usage attribution variable, as described in its attribution reference.
Are the default alert thresholds safe for production?
They are generic defaults based on SRE practice, not tuned to your traffic. Validate them against real traffic and adjust them before relying on them in production.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits platform or SRE engineers, and teams that manage monitoring as Terraform, who need a reviewed baseline of Cloud Run alert policies quickly. It is a starting point, not a finished alerting setup, because its default thresholds are generic. If you run GKE or Compute Engine workloads, or need alerts tuned to your real traffic from day one, pick a different tool or use the GKE skill instead.
Download the skill