What problem does it solve?
This Skill prevents AI coding agents from underperforming due to missing, stale, or overgrown context by systematically curating what the agent sees and how it’s structured.
Core Features & Use Cases
- Context hierarchy control: Loads the most persistent rules files first, then relevant memory, architecture/spec sections, task-relevant source, and finally live state and conversation.
- Drift and quality guardrails: Recommends reloading rules + memory and re-scoping when output quality drifts or when switching between unrelated parts of the codebase.
- High-stakes subagent fan-out: Enforces bounded wave sizes, strict per-agent output budgets, scope bounding, and a required status protocol to avoid context flooding.
- Confusion management pattern: Forces explicit surfacing of conflicting or missing requirements instead of silently choosing a direction.
- Inline planning discipline: Encourages emitting a lightweight plan before executing multi-step changes to reduce rework.
Quick Start
Instruct your agent to apply context engineering by loading the correct rules files and reviewing relevant memory before working on a specific task, then scope any subagent fan-out to a small wave with bounded prompts and outputs.
