subagent-strategy

To prevent context overflow during large-scale document analysis, delegate long-running read and search tasks to subagents. This keeps chapter content and file search results out of the main-thread context window.

Subagent delegation for multi-file analysis works by pushing read_file and search tasks to subagents via the task tool. This enforces hard rules that keep heavy reading off the main thread while retaining edits and writes locally.

Yes, you can batch read multiple chapters without overwhelming the main agent by initiating a task that delegates grouped chapter reads to a subagent. This prevents context bloat by offloading the reading process.

The best way to manage context during multi-dimensional analysis across components is parallel task decomposition. Delegate grouped reads and searches to subagents so the main thread stays clear of bulky content.

Subagent task delegation supports parallel task decomposition for large document searches. By grouping multi-file searches and pushing them to subagents, you execute parallel lookups without filling the main context window.

You should not use subagent delegation for reading tasks that involve direct edits or writes. The workflow keeps edits and writes on the main thread, only offloading long-running lookups and multi-file reads to subagents.