write-agent

Configure ephemeral sub-agents with AGENT.md files for task delegation.

1|Updated Feb 7, 2026
One-click install
npx skills add https://github.com/thirteen37/Krill --skill write-agent-thirteen37
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: write-agent
Source: https://github.com/thirteen37/Krill/tree/main/.claude/plugins/krill-dev/skills/write-agent
Command: npx skills add https://github.com/thirteen37/Krill --skill write-agent-thirteen37

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill streamlines the creation of ephemeral sub-agents, allowing you to delegate specialized tasks to cost-optimized, tool-restricted assistants without cluttering your main agent's context.

Core Features & Use Cases

  • Agent Architecture: Provides a standardized framework for defining agent roles, toolsets, and system prompts via AGENT.md files.
  • Task Delegation: Enables the main agent to spawn sub-agents for specific domains like research, coding, or service configuration.
  • Use Case: If you need to perform deep web research, use this skill to define a research agent that uses cheaper models and specific web-browsing tools, keeping your main agent focused on high-level strategy.

Quick Start

Use the write-agent skill to help me create a new sub-agent for summarizing technical documentation.

Frequently Asked Questions about write-agent

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

FAQPage Schema
How do I delegate tasks to sub-agents without cluttering my main agent's context?

You can delegate tasks to sub-agents by defining specialized agent roles and toolsets via AGENT.md files. This keeps your main agent focused on high-level strategy while ephemeral sub-agents handle domain-specific work.

What is single-level delegation in task automation workflows?

Single-level delegation is a security constraint that ensures spawned sub-agents cannot delegate tasks further. It enforces memory isolation and restricts operational depth for all ephemeral sub-agents within the ecosystem.

How do I configure tool-gating and model-specific performance tuning for automated agents?

You configure tool-gating and model-specific performance tuning by defining specialized agent roles and restricted toolsets within AGENT.md files, ensuring cost-optimized assistants only access designated operational tools.

Can I use cheaper models for deep web research while keeping my main agent focused?

Yes, you can spawn a research sub-agent using cheaper models and specific web-browsing tools. This delegates the research domain while maintaining memory isolation from your main agent's context.

What are the limitations of using ephemeral sub-agents for task delegation?

Ephemeral sub-agents are limited by enforced single-level delegation and memory isolation. They cannot spawn further sub-agents and operate with tool-restricted access, preventing nested workflow chains.