Guide · Agent skill

How to use the Alerting & On-Call Setup agent skill

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

The alerting-oncall skill gives a coding agent a reference guide for setting up production alerts and on-call rotations. It covers severity levels, Prometheus alert rules, Alertmanager routing to Slack or PagerDuty, and PagerDuty services, escalation policies and weekly schedules written in Terraform. This guide explains what the skill does, how to install it, how to ask for help with natural requests, and what to check before you use any output.

Alerting & On-Call Setupsickn33/agentic-awesome-skills · MIT Download the skill

Quick start

1Download the ZIP. Get the alerting-oncall ZIP from the directory page.
2Choose an install location. Unzip to ~/.claude/skills/ for personal use, or to a project's .claude/skills/ folder, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
3Describe your service. Tell the agent the service, monitoring tool, on-call platform, notification channels and time zone.
4Ask for a specific output. Request Prometheus rules, Alertmanager routes or PagerDuty Terraform for one part of the setup at a time.
5Replace placeholders. Swap the xxx keys, example.com runbook links and dashboard URLs for your real values.
6Test in a safe environment. Validate routing and escalation against a non-production receiver before relying on it for real pages.

What the skill makes the agent do

An agent skill is a folder of instructions that an AI agent reads when your request matches the skill's purpose. With this one installed, the agent draws on a guide for configuring alerting and on-call management for production systems.

The skill covers four areas: ranking alert severity (critical, high, medium, low, each with a suggested response), writing alert design rules, producing Prometheus alert rules and Alertmanager configuration, and producing PagerDuty Terraform for services, escalation policies and weekly on-call schedules. It also includes a Grafana OnCall example and guidance on reducing alert fatigue.

The skill only produces configuration text. It does not connect to your monitoring or paging systems and cannot run anything for you.

Severity levels: critical (page immediately), high (page in business hours, notify after hours), medium (Slack, review next business day), low (email, weekly review)
Alert design: actionable, relevant (symptoms over causes), timely and unique
Prometheus rules, Alertmanager routes and receivers, PagerDuty Terraform resources
A reference file on remediation and escalation steps

Install the alerting-oncall skill

Download the ZIP for the skill from the directory page. Then choose one of these options.

For a personal install that applies across your projects, unzip the download into your skills folder at ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. The skill folder should contain SKILL.md and a references folder with details.md.

In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

Personal: ~/.claude/skills/
Project only: .claude/skills/ inside the project
Claude apps: Settings → Capabilities → Skills, then upload

Trigger the skill with natural requests

You do not need to name the skill. Describe the task in plain terms that match its scope, such as setting up alerts for a service or routing alerts to a team. The agent should pick the skill up when your request fits.

Give the agent your context: the service name, what you monitor, your alerting tool, your on-call platform and your time zone. The more specific the request, the more usable the output.

Write Prometheus alert rules for my API that fire on high 5xx error rate and high p95 latency.
Set up Alertmanager so critical alerts page PagerDuty and high alerts go to a Slack channel during business hours.
Create Terraform for a PagerDuty service with a primary weekly rotation and a manager escalation after 15 minutes.
Review these alert rules and tell me which ones are not actionable or are alerting on causes instead of symptoms.

Example: a critical service-down alert

The skill's Prometheus example fires when a target is down for one minute, labels it critical and links to a runbook. Each alert in the guide includes a summary, a description and a runbook_url annotation, which reflects the source's principle that every alert should require action and link to remediation steps.

Replace the example runbook address with your own. The source uses a placeholder domain, so the link will not work as written.

- alert: ServiceDown
  expr: up{job="myapp"} == 0
  for: 1m
  labels:
    severity: critical
  annotations:
    runbook_url: "https://your-wiki.example/runbooks/service-down"

Check and replace placeholders before you apply anything

The example configuration contains placeholder values. The Slack webhook path and the PagerDuty service key are both set to xxx, and the runbook links point to example.com. You must replace each of these with your real values before any configuration is loaded.

Routing mistakes can silence real pages or send them to the wrong team. Test routes against a non-production receiver first, and have someone familiar with your on-call setup review the final result.

Replace the Slack webhook (slack_api_url) and PagerDuty service_key values
Replace example.com runbook and dashboard links with real ones
Confirm that the severity labels in your rules match the routes in Alertmanager
Review escalation delays and the time zone in the PagerDuty schedule

Safety notes: scripts, network and remediation

The skill does not run scripts and does not use the network. It is a documentation-style guide. The compatibility note lists kubectl, helm, git and CI runners, but the main file does not use them. The reference file references/details.md includes one kubectl rollback command for a recent deployment problem. Run it only if you are authorized for that environment and have confirmed a deployment caused the issue.

For database problems, the reference file says not to delete database pods as a general fix. Investigate using the database's own runbook and escalate to the database owner before any recovery step.

No scripts are bundled, and the skill makes no network calls
The kubectl rollback example is in references/details.md; review it before running
Do not follow generic database restart advice without the database owner's approval

kubectl rollout undo deployment/myapp -n production

What people use it for

Writing Prometheus alert rules for a web API covering downtime, error rate and latency
Setting Alertmanager routes so critical alerts page on-call and lower severities go to Slack
Defining a PagerDuty service, escalation policy and weekly primary rotation in Terraform
Reviewing an existing alert list to remove duplicates and alerts that do not need action
Drafting an on-call escalation structure to move a team away from noisy alerts

Frequently asked questions

How do I install the alerting-oncall agent skill?

Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills instead.

What do I need to use the alerting-oncall skill?

You need a monitoring system such as Prometheus or Datadog, an on-call platform such as PagerDuty, Opsgenie or Grafana OnCall, and Slack or email for notifications. Terraform is needed for the PagerDuty examples. The skill only generates text, so you must apply the configuration yourself.

Is the alerting-oncall skill free to use?

The skill is listed under the MIT license, as shown on its page. The license name is the only licence detail in the source, so check the license file linked from the skill's source for full terms.

Can the skill configure PagerDuty or Slack for me?

No. The skill produces configuration examples and guidance but does not connect to PagerDuty, Slack or your monitoring tools. You need to review, adapt and apply the output in your own systems.

Why does the example config not work when I paste it in?

The examples contain placeholders such as a Slack webhook marked xxx, a PagerDuty service key of xxx and example.com runbook links. Replace these with your real values, and check routing before relying on it for real alerts.

How does this compare with writing alert rules by hand?

The skill gives you a starting structure based on a severity model, alert design principles and working examples for Prometheus, Alertmanager and PagerDuty. It does not know your services, thresholds or team structure, so you still need to tune and test everything yourself.

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

Use this skill if you are an engineer or SRE who already runs Prometheus or a similar tool and wants a clear starting point for severity levels, alert routing and PagerDuty escalation. Pick something else if you do not have access to your monitoring or paging systems, or if you need a tool that deploys and tests configuration for you, since this skill only produces text that you must review and adapt.

Download the skill