ki-delegate

Automates decomposition, classification, assignment, sequencing and gating of tasks across agent and model tiers.

Updated May 29, 2026

Search Tags:#task decomposition #agent coordination #multi-tasking #model tiering #safety review

Ready to run

npx skills add https://github.com/knowledgeislands/ki-agentic-harness --skill ki-delegate

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: ki-delegate
Source: https://github.com/knowledgeislands/ki-agentic-harness/tree/main/skills/ki-delegate
Command: npx skills add https://github.com/knowledgeislands/ki-agentic-harness --skill ki-delegate

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill simplifies complex multi-task efforts by distributing tasks across agent and model tiers, ensuring efficient and safe execution.

Core Features & Use Cases

Task Classification: Automatically categorizes tasks as judgment, mechanical, or research.
Agent and Tier Assignment: Matches tasks to appropriate agent types and model tiers based on complexity.
Sequencing and Parallelization: Organizes tasks into rounds, ensuring smooth parallel processing.
Result Review and Gatekeeping: Reviews all results to maintain quality and safety.
Use Case: Ideal for coordinating complex projects with multiple dependent tasks across various agents and models.

Quick Start

Delegate your tasks with "ki-delegate [task-list]".

Frequently Asked Questions about ki-delegate

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

FAQPage Schema

How do I automate task decomposition across multiple agents and model tiers?▼

Task decomposition across agents and model tiers is automated by classifying tasks as judgment, mechanical, or research, then matching them to appropriate models. This ensures complex multi-task efforts are distributed efficiently and safely.

What is the best way to coordinate dependent tasks for parallel processing?▼

Coordinating dependent tasks for parallel processing requires organizing tasks into logical rounds. The sequencing mechanism handles assignment and gating, ensuring smooth execution and safety review across all agents involved.

Do I need specific agents to handle multi-task assignment and execution?▼

Yes, multi-task assignment requires appropriate agents and models for task handling. The system matches task complexity to specific agent types and model tiers, ensuring optimal execution and safety review within the Claude Code runtime.

How does model tiering work for classifying complex project tasks?▼

Model tiering for complex project tasks works by categorizing tasks as judgment, mechanical, or research. The system then matches these classifications to appropriate model tiers based on complexity to optimize execution and safety.

Can I use this to gate and review results across multiple agents?▼

Yes, you can gate and review results across multiple agents. The system manages result review and gatekeeping to maintain quality and safety, ensuring all outputs meet standards before proceeding to the next task round.

Why does task sequencing fail when parallelizing multi-task efforts?▼

Task sequencing fails during multi-task parallelization if tasks are not properly organized into rounds or lack appropriate gating. The system requires correct classification, assignment, and safety review configurations to ensure smooth parallel processing.