Skip to main content
Like AOF? Give us a star!
If you find AOF useful, please star us on GitHub. It helps us reach more developers and grow the community.

Agent Library

AOF includes a production-ready library of 30+ pre-built agents organized by operational domain. These agents are designed by DevOps and SRE experts to solve common operational challenges.

Quick Start​

# List available agents
aofctl get agents --library

# Run an agent from the library
aofctl run agent library://kubernetes/pod-doctor

# Run with custom prompt
aofctl run agent library://kubernetes/pod-doctor \
  --prompt "Debug CrashLoopBackOff in namespace production"

Agent Domains​

DomainAgentsFocus Area
Kubernetes5Pod debugging, HPA tuning, Network policies
Observability7Alerts, SLOs, Dashboards, Logs, Traces, New Relic, Splunk
Incident5RCA, Postmortems, Runbooks, Escalation
CI/CD5Pipelines, Tests, Builds, Releases
Security5Scanning, Compliance, Secrets, Threats
Cloud5Cost, IAM, Right-sizing, Drift
ITSM3ServiceNow incidents, CMDB, Change management

Using Library Agents​

Direct Execution​

# Run with default prompt
aofctl run agent library://observability/log-analyzer

# Run with custom prompt
aofctl run agent library://incident/rca-agent \
  --prompt "Investigate the API latency spike at 14:30 UTC"

# Run interactively
aofctl run agent library://kubernetes/pod-doctor --interactive

Extend with Custom Agents​

Copy a library agent as a starting point:

# Copy agent to local file
aofctl get agent library://kubernetes/pod-doctor -o yaml > my-pod-doctor.yaml

# Customize and run
aofctl run agent -f my-pod-doctor.yaml

Use in Fleets​

Reference library agents in fleet definitions:

apiVersion: aof.sh/v1alpha1
kind: Fleet
metadata:
  name: incident-response
spec:
  agents:
    - ref: library://incident/incident-commander
    - ref: library://incident/rca-agent
    - ref: library://observability/log-analyzer
  strategy:
    type: sequential

Agent Configuration​

All library agents use a consistent configuration:

  • Model: google:gemini-2.5-flash (optimized for DevOps tasks)
  • Max Tokens: 8192
  • Temperature: 0.1-0.3 (low for operational reliability)

Override defaults in your agent definition:

apiVersion: aof.sh/v1alpha1
kind: Agent
metadata:
  name: my-custom-agent
spec:
  # Inherit from library agent
  extends: library://kubernetes/pod-doctor

  # Override settings
  model: anthropic:claude-3-sonnet
  temperature: 0.1

Environment Variables​

Library agents may require environment variables for tool access:

# Kubernetes (uses kubeconfig)
export KUBECONFIG=~/.kube/config

# Observability
export GRAFANA_URL=https://grafana.example.com
export GRAFANA_TOKEN=your-token
export PROMETHEUS_URL=https://prometheus.example.com

# CI/CD
export GITHUB_TOKEN=ghp_xxxx
export GITLAB_TOKEN=glpat-xxxx

# Cloud
export AWS_PROFILE=production
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/credentials.json

Best Practices​

  1. Start with library agents - They encode expert knowledge
  2. Customize prompts - Tailor to your specific environment
  3. Extend, don't rewrite - Use extends for modifications
  4. Combine in fleets - Orchestrate multiple agents for complex tasks
  5. Monitor execution - Review agent decisions in audit logs

Contributing Agents​

Submit new agents to the library:

  1. Follow the Agent Specification
  2. Include comprehensive system prompts
  3. Document required tools and environment variables
  4. Add examples and use cases
  5. Submit PR to the AOF repository

Next Steps​