pool

Execute subagent tasks under a defined concurrency cap.

113|23|Updated Mar 23, 2026
One-click install
npx skills add https://github.com/opensage-agent/opensage-adk --skill pool-opensage-agent
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pool
Source: https://github.com/opensage-agent/opensage-adk/tree/main/src/opensage/bash_tools/workflow/pool
Command: npx skills add https://github.com/opensage-agent/opensage-adk --skill pool-opensage-agent

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the issue of hitting provider rate limits (RPM/TPM) or memory constraints when executing a large batch of subagent tasks simultaneously.

Core Features & Use Cases

  • Sliding-Window Concurrency: Maintains a controlled number of active subagents at any given time.
  • Asynchronous Orchestration: Efficiently polls for task completion without blocking the main thread.
  • Use Case: When running a batch evaluation of 50 independent code review tasks, this Skill ensures only a safe number of agents run in parallel, preventing API throttling and system overload.

Quick Start

Use the pool skill to execute the provided list of tasks with a concurrency limit of six active workers.

Frequently Asked Questions about pool

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent rate limit exhaustion when running batch subagent tasks?

Running 50 independent code review tasks in parallel requires asynchronous orchestration with a defined concurrency limit. This approach efficiently polls for task completion without blocking the main thread, ensuring only a safe number of agents run concurrently to prevent API throttling.

What is the best way to orchestrate asynchronous subagent completion polling?

Asynchronous subagent completion polling requires a session-based subagent management system to track task states. This mechanism efficiently polls for task completion without blocking the main thread, handling asynchronous completion for batch processing scenarios like fan-out explorations.

How do I execute bulk code reviews without blocking the main thread?

Executing bulk code reviews without blocking the main thread involves asynchronous orchestration under a concurrency cap. By polling for task completion asynchronously, the system prevents provider rate limit exhaustion and manages active subagents safely.

Does this concurrency orchestration work for large-scale evaluation benchmarks?

Concurrency orchestration for batch processing requires a session-based subagent management system to track task states. This setup handles asynchronous completion polling and enforces a concurrency cap, such as limiting active workers to six, to prevent provider rate limit exhaustion.