filesystem-context

What problem does it solve?

The Filesystem-Based Context Engineering skill provides a persistent, on-demand storage layer that lets agents offload large tool outputs, scratch work, and long-form notes to the filesystem, reducing prompt size while keeping data accessible on demand.

Core Features & Use Cases

• Scratch pads for large tool outputs and intermediate results with lightweight file references for context.
• Plan persistence to store and rehydrate long-horizon workflows across sessions.
• Sub-agent workspaces for isolated collaboration and coordinated synthesis.
• Dynamic skill loading by keeping skills as files and loading only relevant content on demand.
• Use Case: An agent processes a lengthy dataset by offloading heavy results to scratch files and retrieving only necessary slices during a conversation.

Quick Start

To begin, create a scratch pad and a simple plan, then write a small tool output to scratch and retrieve a reference, demonstrating how large tool outputs are offloaded and referenced in context.