---
schema_version: "newruntime-topic-hub-v0.1"
type: "topic_hub"
slug: "context-engineering"
title: "Context engineering - New Runtime"
description: "How teams package memory, rules, traces, tools, and task state so agents can work without restarting from raw chat context."
answer: ["Context engineering is the work of turning scattered organizational knowledge into durable inputs an agent can inspect and act on.","The strongest systems separate reusable rules, task state, source evidence, and operator corrections instead of burying everything in one prompt.","The practical payoff is lower context waste and more reproducible agent behavior across long-running tasks."]
search_intents: ["context engineering","agent context architecture","LLM context systems"]
status: "featured"
last_updated: "2026-07-21"
counts: {"total":16,"signals":9,"patterns":1,"posts":3,"atlas":3,"sources":27}
routes: {"html":"https://newruntime.com/topics/context-engineering/","markdown":"https://newruntime.com/topics/context-engineering.md","json":"https://newruntime.com/topics/context-engineering.json"}
top_sources: ["https://aclanthology.org/2024.acl-long.747","https://aihero.dev/how-to-kill-the-bloat-in-claude-codes-system-prompt","https://arxiv.org/abs/2410.10813","https://arxiv.org/abs/2602.12670","https://arxiv.org/abs/2603.29919","https://arxiv.org/abs/2605.24050","https://claude.com/blog/steering-claude-code-skills-hooks-rules-subagents-and-more","https://code.claude.com/docs/en/sub-agents","https://dropbox.tech/security/dropbox-mcp-dash-design-code-security","https://engineering.atspotify.com/2026/6/encoding-your-domain-expert-the-context-layer-behind-spotifys-data-assistant","https://engineering.grab.com/scaling-out-distroless-adoption-with-ai","https://github.com/AI-Builder-Club/skills"]
---

# Context engineering - New Runtime

How teams package memory, rules, traces, tools, and task state so agents can work without restarting from raw chat context.

## Short Answer

- Context engineering is the work of turning scattered organizational knowledge into durable inputs an agent can inspect and act on.
- The strongest systems separate reusable rules, task state, source evidence, and operator corrections instead of burying everything in one prompt.
- The practical payoff is lower context waste and more reproducible agent behavior across long-running tasks.

## Patterns

- [Skills become a portable capability layer](https://newruntime.com/patterns/skills-become-portable-capability-layer/): Agent skills are emerging as a portable capability layer, but their value depends on progressive disclosure, provenance, security review, and behavioral evaluation.

## Field Notes

- [AI-Native Engineering Teams Are Smaller and Stricter](https://newruntime.com/posts/ai-native-engineering-team/): AI-native engineering changes team size, roles, onboarding, metrics, and review: code is no longer written only by humans, but context and quality ownership get stricter.
- [Claude Code Subagents Split Memory Instead of Inflating One Session](https://newruntime.com/posts/claude-code-subagents-memory/): Claude Code subagents show a practical form of agent memory hygiene: move noisy work into a separate context and return only the compressed result to the main session.
- [Unabyss Builds a Shared Context Layer Across Claude, Codex, and Cursor](https://newruntime.com/posts/unabyss-shared-context-layer/): Unabyss points to where agent tooling is going: context stops being one client's internal memory and becomes a separate product layer with governance and portability.

## Recent Raw Signals

- 2026-07-16: [Agent skills need behavioral evals, not prose review](https://newruntime.com/signals/agent-skills-need-behavioral-evals/)
- 2026-07-15: [Retrieval and long context do not create company memory](https://newruntime.com/signals/retrieval-does-not-create-company-memory/)
- 2026-07-09: [Claude Code Startup Context Can Be Reduced to 13K Tokens](https://newruntime.com/signals/claude-code-startup-context-can-be-reduced-to-13k-tokens/)
- 2026-07-03: [Codebase Memory Combines Search with a Repository Map](https://newruntime.com/signals/codebase-memory-combines-search-with-a-repository-map/)
- 2026-06-28: [Pydantic AI loads capabilities only when needed](https://newruntime.com/signals/pydantic-ai-capabilities-progressive-disclosure/)
- 2026-06-23: [CLAUDE.md Shrinks as Skills, Hooks, and Subagents Mature](https://newruntime.com/signals/claude-md-shrinks-as-skills-hooks-and-subagents-mature/)
- 2026-06-16: [Production AI Engineering Is the System Around the Model](https://newruntime.com/signals/production-ai-engineering-is-the-system-around-the-model/)
- 2026-05-24: [Codex Subagents Reduce Context Pollution](https://newruntime.com/signals/codex-subagents-reduce-context-pollution/)
- 2026-05-07: [SKILL.md Acts as a Behavior Loader, Not a Better Prompt](https://newruntime.com/signals/skill-md-acts-as-a-behavior-loader-not-a-better-prompt/)

## Public Sources

- https://aclanthology.org/2024.acl-long.747
- https://aihero.dev/how-to-kill-the-bloat-in-claude-codes-system-prompt
- https://arxiv.org/abs/2410.10813
- https://arxiv.org/abs/2602.12670
- https://arxiv.org/abs/2603.29919
- https://arxiv.org/abs/2605.24050
- https://claude.com/blog/steering-claude-code-skills-hooks-rules-subagents-and-more
- https://code.claude.com/docs/en/sub-agents
- https://dropbox.tech/security/dropbox-mcp-dash-design-code-security
- https://engineering.atspotify.com/2026/6/encoding-your-domain-expert-the-context-layer-behind-spotifys-data-assistant
- https://engineering.grab.com/scaling-out-distroless-adoption-with-ai
- https://github.com/AI-Builder-Club/skills