prometheus-observer

I analyze running Prometheus installations to provide comprehensive reports on current alert states, configuration health, and system status without making any modifications.

Slash Command

/prometheus-status

Runs the full autonomous validation workflow:

1. Verify Kubernetes cluster connectivity
2. Check Prometheus and AlertManager pods exist
3. Test port-forwarded endpoints (localhost:9090, localhost:9093)
4. Query active alerts and report status
5. Check scrape target health

Usage: Just type /prometheus-status and I will execute the validation script and report results.

Script Verification: Before executing, verify the script integrity:

sha256sum .github/skills/prometheus-observer/scripts/validate.sh
# Expected: check current hash after creation

Execute validation:

bash .github/skills/prometheus-observer/scripts/validate.sh

When I Activate

• /prometheus-status (slash command)
• "Is Prometheus running?"
• "Check Prometheus status"
• "What alerts are firing?"
• "Validate AlertManager config"
• "Show Prometheus rules"
• "Prometheus health check"
• "Alert analysis"

Port-Forward Assumptions

This skill assumes port-forwards are active or can be started:

• Prometheus: kubectl port-forward -n monitoring svc/kube-prometheus-stack-prometheus 9090:9090 &
• AlertManager: kubectl port-forward -n monitoring svc/kube-prometheus-stack-alertmanager 9093:9093 &

If port-forwards are not running, the validation script will detect this and provide the commands to start them.

Core Capabilities

1. Alert State Analysis

• Query Prometheus API (/api/v1/alerts) for active/pending alerts
• Categorize alerts by severity and state
• Report alert duration and frequency patterns
• Identify which rules are triggering most frequently

2. AlertManager Integration Status

• Verify AlertManager connectivity via Prometheus config
• Check AlertManager API health (/-/healthy)
• Validate notification routing configuration
• Report on silenced alerts and inhibition rules

3. Rule Evaluation Health

• Query rule evaluation status (/api/v1/rules)
• Identify failed rule evaluations
• Report on rule evaluation latency
• Check for missing metrics in rule queries

4. Configuration Validation

• Parse Prometheus configuration for syntax issues
• Validate scrape target connectivity
• Check for missing labels or misconfigured jobs
• Report on retention and storage settings

Usage Examples

Basic Status Check

# Query active alerts
curl -s http://prometheus:9090/api/v1/alerts | jq '.data.alerts[] | {alert: .labels.alertname, state: .state, severity: .labels.severity}'

# Check AlertManager connectivity  
curl -s http://alertmanager:9093/-/healthy

Advanced Analysis

# Rule evaluation metrics
curl -s http://prometheus:9090/api/v1/query?query=prometheus_rule_evaluation_failures_total | jq '.data.result[]'

# Scrape health by job
curl -s http://prometheus:9090/api/v1/query?query=up | jq '.data.result[] | {job: .metric.job, instance: .metric.instance, status: .value[1]}'

Common API Endpoints

| Endpoint | Purpose | |----------|---------| | /api/v1/alerts | Active and pending alerts | | /api/v1/rules | Rule evaluation status | | /api/v1/targets | Scrape target health | | /api/v1/config | Current configuration | | /-/healthy | Health status |

Monitoring Patterns

Application-Specific Alerts

• Application readiness monitoring
• Deployment failure detection
• Component availability tracking

AlertManager Routing Validation

# Example route structure to validate
route:
  group_by: ['alertname']
  group_wait: 10s
  group_interval: 10s
  repeat_interval: 1h
  receiver: 'web.hook'

Troubleshooting Guide

No Alerts Firing

• Check rule evaluation: are rules syntactically correct?
• Verify metrics availability: are scraped targets up?
• Validate time windows: are alert conditions realistic?

AlertManager Not Receiving Alerts

• Confirm Prometheus → AlertManager configuration
• Check AlertManager logs for webhook errors
• Validate notification receiver configuration

High Rule Evaluation Latency

• Review rule complexity and query performance
• Check for missing indices on high-cardinality metrics
• Validate scrape interval vs evaluation interval

MCP Server Accelerators

Want faster, more comprehensive analysis? These MCP servers provide deeper integration than manual API queries:

Flux Operator MCP (Recommended First)

The Flux Operator MCP Server (flux-operator) provides Kubernetes resource access that complements Prometheus observation:

• Use get_kubernetes_resources to query PrometheusRule and ServiceMonitor objects
• Access pod logs for Prometheus/AlertManager directly
• Combined with this skill's API queries, gives complete monitoring picture

Setup: See flux-operator skill Step 10 for MCP server configuration.

Prometheus MCP Server

For direct Prometheus API integration, consider:

• prometheus-mcp-server (Golang) - Full API support, STDIO/SSE/HTTP transports
• prometheus-mcp-server (TypeScript) - Natural language queries

These provide tool-based access to the same APIs this skill uses manually.

Grafana MCP Server

If you're using Grafana for visualization:

• mcp-grafana - Search dashboards, investigate incidents, query datasources

Integration Points

This skill provides the observational foundation for:

• Flux Operator → Use MCP server for PrometheusRule resource access
• AlertManager Installer - Understanding current state before changes
• KSM Crossplane Adapter - Validating new metrics are collected
• Resource Template Engine - Confirming new alerts activate properly