Obol Agent Skills

A Claude Code plugin with skills for running decentralised infrastructure with Obol.

Installation

As a Claude Code Plugin (Recommended)

/plugin marketplace add ObolNetwork/skills
/plugin install obol

Reload plugins if needed:

/reload-plugins

Manual Installation

Copy the skills/ directory into your project's .claude/skills/ folder.

Configuration

Required for obol-monitoring skill: Grafana API Token

Set the OBOL_GRAFANA_API_TOKEN environment variable. The Obol Core team can provide you with one.

Option 1 — Shell profile (recommended):

# Add to ~/.bashrc or ~/.zshrc
export OBOL_GRAFANA_API_TOKEN="glsa_..."

Option 2 — Project .env file (see .env.sample):

OBOL_GRAFANA_API_TOKEN=glsa_...

Available Skills

obol-monitoring

Monitor and diagnose Obol DVT cluster performance using Grafana (Prometheus metrics + Loki logs).

Scripts:

• cluster_triage.py — First-pass cluster health check
• duty_analysis.py — Deep slot-level failure analysis with timeline reconstruction
• fleet_overview.py — Multi-cluster fleet view with version/client diversity

Usage: Ask Claude to triage a cluster, analyze duty failures, or get a fleet overview. The skill guides Claude through the appropriate diagnostic workflow.

See skills/obol-monitoring/SKILL.md for the full reference including failure reason codes, metrics guide, and triage workflow.

Requirements

• Python 3.6+ (stdlib only, no pip packages needed)
• OBOL_GRAFANA_API_TOKEN environment variable

Adding New Skills

Create a new directory under skills/ with a SKILL.md file:

skills/
├── obol-monitoring/
│   ├── SKILL.md
│   └── scripts/
└── your-new-skill/
    ├── SKILL.md
    └── ...