agent-context-isolation

Prevent agent output from polluting the main context window. Background agents with run_in_background=true isolate their context. Have them write results to files in .claude/cache/agents/ / . TaskOutput returns the full agent transcript. Always use file-based coordination instead. Stuck agent detection: After agent work: Each agent reads the previous agent's file output, not TaskOutput. Agent context isolation preserves the main conversation's context budget. Reading agent outputs via TaskOutput floods context, causing: - Progress reminders stop arriving - Task output file size stops growing - Expected output file not created after reasonable time - Run the test suite directly: bun test - Report pass/fail counts - Only investigate failures if tests fail - Mid-conversation compaction - Lost context about user's original request - Repeated explanations needed - Session where TaskOutput flooded 70k+ tokens into main context - Session 2026-01-01: Successfully used background agents with file-based coordination for SDK Phase 3