managing-agents

Agent lifecycle management coordinates creation, execution, promotion, and automatic cleanup of all agent types—built-in, plugin-defined, and temporary. This ensures your AI toolkit remains organized and removes obsolete agents automatically, preventing clutter and redundancy.

Temporary agents are workflow-specific, short-lived instances created for immediate tasks. Permanent agents are reusable assets registered in the agent registry. You can promote successful temporary agents to permanent status, capturing valuable expertise for future workflows.

Promotion moves a temporary agent into your permanent registry via filesystem-based workflows. After validation, the agent becomes available organization-wide with namespace prefixing applied, eliminating the need to redefine it for each new workflow.

Namespacing ensures unique agent identification through automatic prefixing and registry routing rules. This prevents naming collisions, improves discoverability, and enforces consistent routing across your AI ecosystem without manual conflict resolution.

Yes, the lifecycle framework handles all three agent types—built-in, plugin-defined, and temporary—uniformly through agents/registry.json. Type enumeration and namespace rules apply consistently, streamlining management across heterogeneous agent sources.

Automatic cleanup removes temporary and obsolete agents based on lifecycle policies, preventing registry bloat. Permanent agents persist unless explicitly removed, giving you control over what stays in your active toolkit.