What kind of session anomalies can I track?

You can monitor for long session durations (e.g., >1 hour), high turn counts, excessive subagent spawning, and high error rates per session.

How do I ensure my alert queries are working correctly?

The skill includes a 'test-alert' operation that allows you to simulate alert conditions and verify your LogQL queries before deployment.

Does this skill help with context window management?

Absolutely. It includes templates to alert you when context utilization exceeds 80% or when auto-compaction is triggered.

Can I receive alerts on Slack or PagerDuty?

Yes, the skill supports multiple notification channels including Slack webhooks, Email, PagerDuty, and Grafana dashboard annotations.

What data source does this skill use for monitoring?

The Observability Alert Manager primarily uses Loki logs, specifically targeting the 'claude_code_enhanced' job telemetry.

Observability Alert Manager

Name: Observability Alert Manager
Author: adaptationio

byadaptationio

0•

Analytics & Monitoring

Configures and manages Grafana alerts for Claude Code to monitor session anomalies, error rates, and resource utilization.

The Observability Alert Manager provides a robust framework for monitoring Claude Code performance and behavior using Grafana and Loki. It allows developers to define, list, test, and manage automated alerts for specific events such as high error rates, context window saturation, long session durations, and unusual subagent spawning. By integrating directly with enhanced telemetry data, it ensures that teams are immediately notified via Slack, email, or PagerDuty when Claude Code encounters operational anomalies or exceeds predefined resource thresholds, helping maintain optimal AI performance and cost control.

Key Features

01Automated LogQL alert generation for Grafana and Loki data sources

020 GitHub stars

03Real-time tracking of context window utilization and auto-compaction triggers

04Comprehensive session monitoring for duration, turn counts, and errors

05Pre-built templates for detecting subagent spikes and unusual activity patterns

06Support for multi-channel notifications including Slack, PagerDuty, and Email

Use Cases

01Detecting and resolving sudden spikes in tool execution errors during automated coding tasks

02Monitoring context window usage to prevent data loss or efficiency drops during complex sessions

03Tracking unexpected increases in subagent spawning to manage API consumption and operational costs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add adaptationio/skrillz observability-alert-manager

For use in Claude.ai and ChatGPT

Download Skill