Agent Skills for Context Engineering

Install

install

$ /plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering

Update

update

$ /plugin marketplace update context-engineering-marketplace

Then run /reload-plugins. Marketplace slug is context-engineering-marketplace, not the repo name.

When to use it

You are building a production agent and need disciplined patterns for managing the context window beyond ad-hoc prompting.
Your agent degrades over long sessions and you need compression, memory, and context-optimization strategies.
You are designing a multi-agent system and want vetted orchestration, tool-design, and filesystem-context patterns.
You need to evaluate or benchmark an agent harness and want rubrics, provenance tracking, and adversarial eval scaffolding.

Commands & flags

/plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering: Register the marketplace in Claude Code.
/plugin install context-engineering@context-engineering-marketplace: Install the full context-engineering plugin (all 15 skills).
context-fundamentals: Skill on managing the model's limited attention budget vs prompt engineering.
context-compression: Skill for compacting context to fit the window without losing signal.
multi-agent-patterns: Skill covering orchestration and coordination across multiple agents.
harness-engineering: Skill for designing the agent harness: tools, memory, eval loops, run state.

How to use it

/plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering
/plugin install context-engineering@context-engineering-marketplace

This loads all 15 skills (context-fundamentals, context-degradation, context-compression, multi-agent-patterns, memory-systems, tool-design, filesystem-context, hosted-agents, context-optimization, latent-briefing, evaluation, advanced-evaluation, harness-engineering, project-development, bdi-mental-states). Claude auto-invokes the relevant skill by task description; you can also name a skill directly. The same skills are exposed for Cursor and other platforms via the Open Plugins directory, and can be dropped into .claude/skills/ standalone.

Why it earns its place

A focused, well-organized reference set codifying context-engineering and harness-engineering as a discipline — context degradation, compression, memory systems, multi-agent coordination, tool design, and evaluation — rather than scattered prompting tips. It ships as a real Claude Code marketplace (plugin v2.3.0) with a parallel Open Plugins manifest, so the same skills work across Claude Code, Cursor, and other conformant agent platforms, and bundles evaluation scaffolding (rubrics, provenance, run-state machines, adversarial benchmarking) for hardening a harness.

Try it

claude

> Use the context-engineering skills to redesign my agent's harness: my long-running agent loses track of earlier decisions — apply context-degradation and memory-systems patterns.