Summary

Separating planning from execution reduces context bloat, increases reliability, and makes handoffs more legible.

Key points

• The planner defines scope, questions, evidence requirements, and stop conditions.
• The executor receives only task-local context and the output contract.
• This separation improves repeatability and reduces contamination from oversized working context.
• It works best when paired with a standard handoff artifact.

Operational relevance

This is one of the most actionable operating-model hardening moves for OpenClaw and similar systems.

Open questions

• Which real workflow should become the first durable planner/executor pilot?
• What is the minimum viable task-card schema?

Sources

• wiki/sources/openclaw-operating-model-gap-assessment.md
• wiki/sources/openai-harness-engineering.md