What is filesystem-based context engineering?

It is a technique where an AI agent uses the local filesystem to store and retrieve data on-demand, allowing it to handle information that exceeds its active context window.

What tools does this skill use for retrieval?

It leverages standard filesystem tools like ls, glob, and grep, along with line-range reads, to discover and pull only the specific data needed for a given task.

Can I use this for multi-agent collaboration?

Yes. This skill enables a 'shared state' pattern where multiple agents can read and write to the same files, preventing the 'game of telephone' information loss common in message passing.

How does this skill help reduce token costs?

By writing large outputs (like logs or search results) to files instead of including them in the chat history, you prevent the prompt from growing unnecessarily large every turn.

Filesystem Context Manager

Name: Filesystem Context Manager
Author: guanyang

byguanyang

•

124

Productivity & Workflow

Manages and offloads agent context to the filesystem to overcome context window limits and improve token efficiency.

About

The Filesystem Context Manager empowers Claude with advanced context engineering capabilities by utilizing the local filesystem as a dynamic memory layer. This skill addresses the fundamental constraint of limited context windows by enabling agents to offload large tool outputs, persist long-term plans, and facilitate high-fidelity communication between sub-agents. By implementing modular patterns like scratch pads and just-in-time discovery, it ensures that only the most relevant information enters the prompt, resulting in higher performance and lower costs for complex, multi-step development tasks.

Key Features

Tool output offloading to prevent prompt bloat from large datasets
Plan persistence for maintaining focus during long-horizon tasks
Targeted information retrieval using grep and line-specific reads
Dynamic 'just-in-time' skill loading to reduce static context overhead
High-fidelity sub-agent coordination via shared filesystem state
124 GitHub stars

Use Cases

Maintaining complex refactoring checklists across multiple conversation turns
Building multi-agent workflows where agents share state via structured workspace files
Processing massive web search or database results without losing chat history

About

Key Features

Tool output offloading to prevent prompt bloat from large datasets
Plan persistence for maintaining focus during long-horizon tasks
Targeted information retrieval using grep and line-specific reads
Dynamic 'just-in-time' skill loading to reduce static context overhead
High-fidelity sub-agent coordination via shared filesystem state
124 GitHub stars

Use Cases

Maintaining complex refactoring checklists across multiple conversation turns
Building multi-agent workflows where agents share state via structured workspace files
Processing massive web search or database results without losing chat history