parallel-agents

To prevent context bloat during parallel agent execution, use file-based status reporting instead of TaskOutput calls to track completion. This approach optimizes token usage and provides structured monitoring for concurrent agent tasks.

You can launch up to 15 agents concurrently for batch processing tasks. This parallel execution capability allows efficient handling of large-scale operations like distributed research or data backfill across multiple providers.

The best way to track parallel agent batch completion status is through a shared status file. Agents running in the background report their completion to this file, enabling efficient monitoring without context bloat.

Yes, you can run agents in the background for distributed research tasks. This Skill supports background execution with batch size limitations up to 15 concurrent agents, reporting completion via file-based confirmations.

Concurrent agents cause token inefficiency when TaskOutput calls accumulate large amounts of context data. Using file-based confirmations instead of direct output calls prevents this context bloat and optimizes overall token usage.