Sókrates Wiki

Independent Agent Team Spawning Process

2 min read

Independent Agent Team Spawning Process

Summary

The process of initializing and coordinating independent agent “teammates” within the development environment to execute parallel tasks across multiple repositories. This workflow favors a flat team structure over a hierarchical subagent model, utilizing a centralized task board to manage dependencies and state.

Details

The Sokrates development lifecycle utilizes an orchestration pattern where the primary agent spawns specialized “teammates” to handle complex, multi-repo engineering tasks. This process is distinct from standard subagent spawning; it emphasizes the independence of the agents, allowing them to operate in parallel while remaining synchronized through a shared task management system.

Workflow Stages

  1. Team Initialization: The process begins with the TeamCreate tool, which establishes the organizational context for the session. This step moves the operation from a single-agent context to a multi-agent collaborative environment.
  2. Task Board Configuration: A series of tasks are defined using the TaskCreate tool. These tasks represent the granular requirements of the feature or fix being implemented. In the documented instance, this involved setting up a task board to track progress across backend and frontend domains.
  3. Dependency Mapping: Using the TaskUpdate tool, the primary agent defines the dependency chain between tasks. This ensures that while agents work independently, they respect the logical order of operations (e.g., ensuring an API endpoint is defined before the frontend attempts to consume it).
  4. Teammate Spawning: Once the environment is prepared, the agent spawns specific teammates. A common configuration involves spawning a backend agent and a frontend agent. These teammates are capable of working across different repositories simultaneously, such as modifying the Eidos knowledge graph API while updating the Hermes Agent channel I/O logic.

Architectural Preference

The source data highlights a specific decision to use “independent team agents” rather than “subagents.” This distinction suggests a preference for agents that have their own context and autonomy within their assigned scope, rather than being strictly controlled step-by-step by a parent process. This approach is designed to increase development velocity by allowing parallel execution in the NixOS-based stack.

Graph View