Prometheus

Query Prometheus monitoring data to check server metrics, resource usage, and system health. Use when the user asks about server status, disk space, CPU/memo...

Security Scan

The skill's code, runtime instructions, and requested resources are consistent with a Prometheus query CLI — nothing indicates covert or unrelated behavior.

Purpose & Capability

Name/description (query Prometheus metrics) align with the files and code: the CLI reads a config or environment, performs Prometheus HTTP API calls, supports multi-instance queries and basic auth — all expected for this purpose.

Instruction Scope

SKILL.md directs running the provided node scripts and storing a config in the OpenClaw workspace; the code also loads optional .env files from the workspace and CWD which could populate process.env. Reading local .env files and allowing PROMETHEUS_* env fallbacks is plausible for convenience, but it means the skill will read local environment files beyond only a single dedicated config file.

Install Mechanism

No install spec; the skill is instruction + Node.js scripts. That is low-risk and consistent with a CLI-style skill.

Credentials

The skill declares no required env vars, and the code only uses PROMETHEUS_URL/USER/PASSWORD and any entries from a local prometheus.json. However, the loader will parse .env files in the workspace and CWD and set any keys not already present in process.env — this can pull unrelated secrets (if present in .env) into process.env for the process, though they are used only as fallbacks and not exfiltrated by the code.

Persistence & Privilege

always is false and the skill writes only its own config (prometheus.json) under the OpenClaw workspace or a path provided by PROMETHEUS_CONFIG. It does not modify other skills or system-wide agent settings.

Assessment

This skill appears to do exactly what it says: query Prometheus HTTP APIs for metrics. Before installing, note that the CLI will read and set environment variables from local .env files (workspace and current directory) and will write a config file (prometheus.json) into your OpenClaw workspace (or the path you provide). Don’t point it at untrusted Prometheus endpoints; review the generated prometheus.json to ensure no sensitive credentials are stored there in plaintext. If you keep secrets in a workspace .env, be aware the CLI may load them into process.env (it does not transmit them anywhere except to the Prometheus instances you configure).

Prometheus Skill

Query Prometheus monitoring data from one or multiple instances. Supports federation across multiple Prometheus servers with a single command.

Quick Start

1. Initial Setup

Run the interactive configuration wizard:

cd ~/.openclaw/workspace/skills/prometheus
node scripts/cli.js init

This will create a prometheus.json config file in your OpenClaw workspace (~/.openclaw/workspace/prometheus.json).

2. Start Querying

# Query default instance
node scripts/cli.js query 'up'

# Query all instances at once
node scripts/cli.js query 'up' --all

# List configured instances
node scripts/cli.js instances

Configuration

Config File Location

By default, the skill looks for config in your OpenClaw workspace:

~/.openclaw/workspace/prometheus.json

Priority order:

1. Path from PROMETHEUS_CONFIG environment variable
2. ~/.openclaw/workspace/prometheus.json
3. ~/.openclaw/workspace/config/prometheus.json
4. ./prometheus.json (current directory)
5. ~/.config/prometheus/config.json

Config Format

Create prometheus.json in your workspace (or use node cli.js init):

{
  "instances": [
    {
      "name": "production",
      "url": "https://prometheus.example.com",
      "user": "admin",
      "password": "secret"
    },
    {
      "name": "staging",
      "url": "http://prometheus-staging:9090"
    }
  ],
  "default": "production"
}

Fields:

• name — unique identifier for the instance
• url — Prometheus server URL
• user / password — optional HTTP Basic Auth credentials
• default — which instance to use when none specified

Environment Variables (Legacy)

For single-instance setups, you can use environment variables:

export PROMETHEUS_URL=https://prometheus.example.com
export PROMETHEUS_USER=admin        # optional
export PROMETHEUS_PASSWORD=secret   # optional

Usage

Global Flags

Flag	Description
-c, --config <path>	Path to config file
-i, --instance <name>	Target specific instance
-a, --all	Query all configured instances

Commands

Setup

# Interactive configuration wizard
node scripts/cli.js init

Query Metrics

cd ~/.openclaw/workspace/skills/prometheus

# Query default instance
node scripts/cli.js query 'up'

# Query specific instance
node scripts/cli.js query 'up' -i staging

# Query ALL instances at once
node scripts/cli.js query 'up' --all

# Custom config file
node scripts/cli.js query 'up' -c /path/to/config.json

Common Queries

Disk space usage:

node scripts/cli.js query '100 - (node_filesystem_avail_bytes / node_filesystem_size_bytes * 100)' --all

CPU usage:

node scripts/cli.js query '100 - (avg by (instance) (irate(node_cpu_seconds_total{mode="idle"}[5m])) * 100)' --all

Memory usage:

node scripts/cli.js query '(node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100' --all

Load average:

node scripts/cli.js query 'node_load1' --all

List Configured Instances

node scripts/cli.js instances

Output:

{
  "default": "production",
  "instances": [
    { "name": "production", "url": "https://prometheus.example.com", "hasAuth": true },
    { "name": "staging", "url": "http://prometheus-staging:9090", "hasAuth": false }
  ]
}

Other Commands

# List all metrics matching pattern
node scripts/cli.js metrics 'node_memory_*'

# Get label names
node scripts/cli.js labels --all

# Get values for a label
node scripts/cli.js label-values instance --all

# Find time series
node scripts/cli.js series '{__name__=~"node_cpu_.*", instance=~".*:9100"}' --all

# Get active alerts
node scripts/cli.js alerts --all

# Get scrape targets
node scripts/cli.js targets --all

Multi-Instance Output Format

When using --all, results include data from all instances:

{
  "resultType": "vector",
  "results": [
    {
      "instance": "production",
      "status": "success",
      "resultType": "vector",
      "result": [...]
    },
    {
      "instance": "staging",
      "status": "success",
      "resultType": "vector",
      "result": [...]
    }
  ]
}

Errors on individual instances don't fail the entire query — they appear with "status": "error" in the results array.

Common Queries Reference

Metric	PromQL Query
Disk free %	node_filesystem_avail_bytes / node_filesystem_size_bytes * 100
Disk used %	100 - (node_filesystem_avail_bytes / node_filesystem_size_bytes * 100)
CPU idle %	avg by (instance) (irate(node_cpu_seconds_total{mode="idle"}[5m])) * 100
Memory used %	(node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100
Network RX	rate(node_network_receive_bytes_total[5m])
Network TX	rate(node_network_transmit_bytes_total[5m])
Uptime	node_time_seconds - node_boot_time_seconds
Service up	up

Notes

• Time range defaults to last 1 hour for instant queries
• Use range queries [5m] for rate calculations
• All queries return JSON with data.result containing the results
• Instance labels typically show host:port format
• When using --all, queries run in parallel for faster results
• Config is stored outside the skill directory so it persists across skill updates