What problem does it solve? LLM context windows are limited, and large tool outputs, long-running plans, and multi-agent communication quickly bloat or overflow the window. This Skill uses the filesystem as an overflow layer so agents store, retrieve, and update context on demand instead of keeping everything in the prompt. ## Core Features & Use Cases - Tool Output Offloading: Write large tool outputs to scratch files and return compact summaries with file references, keeping only ~100 tokens in context. - Plan Persistence: Save structured plans (JSON/YAML) to disk so agents can re-read objectives and progress after context refreshes or summarization. - Sub-Agent Workspaces & Dynamic Skill Loading: Route sub-agent findings through isolated file directories and load full skill files only when relevant, converting O(n) static token cost into O(1) per task. - Use Case: A web search returns 8000 tokens; the agent writes it to scratch/search_results_001.txt, returns a one-line summary, and greps the file later for specific details. ## Quick Start Ask the agent to offload large tool outputs to scratch files and persist the current plan to disk so it can re-read them on demand.
