meow:lazy-agent-loader

Loads agent definitions on-demand to minimize token consumption in Claude Code workflows.

14|2|Updated Mar 25, 2026

Search Tags:#lazy-loading #token-savings #context-budget #on-demand-loading #agent-loader #session-cache #agent-detection

Ready to run

npx skills add https://github.com/ngocsangyem/MeowKit --skill meow-lazy-agent-loader

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: meow:lazy-agent-loader
Source: https://github.com/ngocsangyem/MeowKit/tree/main/.claude/skills/meow%3Alazy-agent-loader
Command: npx skills add https://github.com/ngocsangyem/MeowKit --skill meow-lazy-agent-loader

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This approach reduces token waste by deferring full agent loading until an agent is activated, keeping summaries lightweight at startup.

Core Features & Use Cases

On-demand loading of agent definitions to save tokens during multi-agent workflows.
Summary-first indexing with full definitions loaded only for PRIMARY candidates.
Session-level caching to avoid re-loading and to preserve context across interactions.
Seamless integration with agent-detection to minimize context expansion while preserving accuracy.

Quick Start

Use the lazy-agent-loader to load agents on demand during a session to minimize token usage.

Frequently Asked Questions about meow:lazy-agent-loader

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

FAQPage Schema

How do I reduce token usage when loading multiple agents in Claude Code?▼

You can reduce token usage by implementing lazy loading, which presents lightweight agent summaries at startup and fetches full definitions only when an agent is activated.

What is on-demand agent loading and how does it minimize context budget?▼

On-demand agent loading minimizes context budget by storing summaries initially and loading full agent definitions only for selected candidates, caching results for session continuity.

How do I cache agent definitions for session continuity in multi-agent workflows?▼

You can cache agent definitions for session continuity by loading full profiles upon activation and using a reload mechanism to refresh cached definitions when needed.

Can I selectively load agent definitions for primary candidates only?▼

Yes, selective loading fetches full agent definitions exclusively for primary candidates by indexing summaries first and loading complete profiles upon activation.

Does lazy agent loading work with agent-detection to prevent context expansion?▼

Yes, lazy loading integrates with agent-detection to minimize context expansion while preserving accuracy by deferring full definition loading until activation.

When should I avoid loading all agent definitions at startup?▼

You should avoid loading all definitions at startup during multi-agent workflows when token consumption is a concern, as deferring loading cuts token waste.