Skip to main content

Module core

construct::agent::operator

Module core 

Source
Expand description

Universal operator prompt — provider-agnostic orchestration philosophy.

This is the core layer that works with any LLM. It defines what the operator does (plan, delegate, monitor, synthesize) without prescribing how to format tool calls — that is handled by provider-specific layers.

§Compact-first design (OpenClaw pattern)

The prompt is split into two tiers:

  • OPERATOR_CORE_PROMPT — compact reference (~600 tokens) injected every turn
  • OPERATOR_FULL_REFERENCE — detailed manual (~2,500 tokens) loaded on-demand via load_skill when the agent needs deep orchestration guidance.

Constants§

OPERATOR_CORE_PROMPT
Compact operator instructions injected every turn (~600 tokens).
OPERATOR_FULL_REFERENCE
Detailed operator reference loaded on-demand (~2,500 tokens).