Summary

This skill provides operational guidance for managing long-lived agent workloads in enterprise environments, covering runtime lifecycle (start/pause/stop), observability (logs/metrics/traces), safety controls (scopes/permissions/kill switches), and change management (rollout/rollback/audit). It documents baseline controls like immutable artifacts, least-privilege credentials, and audit logging, along with incident response patterns and integration points with deployment tools.

Detected Capabilities

Trigger Keywords

Phrases that MCP clients use to match this skill to user intent.

Use Cases

• Monitor and manage continuously running agent systems in production
• Implement safety controls and audit trails for multi-agent deployments
• Execute controlled rollout and rollback procedures for agent updates
• Respond to agent failures with structured incident patterns and isolation procedures
• Integrate agent operations with PM2, systemd, container orchestrators, and CI/CD pipelines

Quality Notes

• Skill lacks concrete implementation examples or code patterns — describes concepts but provides no actionable instructions for an agent
• No specific metrics definitions, thresholds, or monitoring setup guidance provided
• Incident response pattern is high-level and generic; lacks specific commands or procedures an agent could follow
• No details on credential management beyond 'least-privilege' and 'environment-level injection'
• Missing guardrails or scope boundaries around which systems/agents can be operated on
• No error handling guidance for failed rollouts, lifecycle operations, or recovery
• Deployment integration section lists tools but provides no instructions for using them
• No edge cases addressed (e.g., what happens if an agent is already stopped before pause, how to handle orphaned processes)
• Very limited scope — skill is mostly conceptual framework rather than actionable operational guidance