spill-context-to-disk

Moves bulky conversation content onto disk and loads back only the slice the current step needs. Covers parking oversized tool results and long command output in timestamped files with short digests, keeping a plan file that is re-read every turn, archiving permitted conversation history so a window reset loses no detail, exchanging findings with other agents through per-agent folders, loading reference documents on demand from a compact index, and storing user preferences across sessions with explicit confirmation. Use when tool output or logs crowd the window, a task spans many turns or must survive a reset or summarization, several agents must share results, or the user asks for file-based memory or on-demand context loading. Not for short tasks whose content already fits, latency-critical work where file round-trips hurt, or building embedding-based semantic search.

Version

v1.0.0

Size

16 KB

Updated

2026-10-11

Your client verifies the signature before reading. This explicit manual read is available on every plan. Free automatically loads five chosen skills; paid plans automatically load relevant skills across the library.