context-engineering

Align AI agent session context with project phases to prevent drift.

6|Updated Mar 30, 2026
One-click install
npx skills add https://github.com/moberghr/mtk-agent-toolkit --skill context-engineering-moberghr
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: context-engineering
Source: https://github.com/moberghr/mtk-agent-toolkit/tree/main/.claude/skills/context-engineering
Command: npx skills add https://github.com/moberghr/mtk-agent-toolkit --skill context-engineering-moberghr

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Guides AI agents to maintain the correct context and phase alignment during coding sessions, preventing drift from project norms.

Core Features & Use Cases

  • Phase-aware context loading and refresh
  • Phase-switch guidance for planning, implementation, and review
  • Reference scoping based on touched files to reduce cognitive load
  • Local pattern anchoring to keep consistent coding styles across tasks

Quick Start

Start a new session with a clearly defined phase and load only the relevant references for the touched files.

Frequently Asked Questions about context-engineering

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent AI agents from drifting from project norms during a coding session?

To prevent context drift in AI coding sessions, align the agent's session context with project phases and enforce reference scoping based on touched files. This maintains consistent, targeted behavior across tasks.

How do I manage context loading when switching between planning, implementation, and review phases?

Manage context loading during phase-switches by applying phase-aware guidance to load only references relevant to the new phase. This reduces cognitive load and ensures the AI follows phase-specific norms.

What is context engineering for AI coding sessions?

Context engineering for AI coding sessions is the practice of aligning AI output with project context. It involves enforcing reference loading rules, scoping, and refresh procedures to prevent output drift.

How do I scope references to reduce cognitive load for an unfamiliar codebase exploration?

Scope references for unfamiliar codebase exploration by loading only the context relevant to the touched files. This enforces targeted reference loading rules, reducing cognitive load and keeping AI behavior reproducible.

When should I refresh AI session context to maintain reproducible behavior?

Refresh AI session context when starting a new session, switching phases, exploring unfamiliar code, or noticing output drift. This enforces context refresh procedures to maintain reproducible, phase-aligned behavior.