Skip to main content
AI Inspiration World
AI AppsAI SkillsAI GamesAI Short FilmsAI Prompts
中文About
Search⌕中
AI AppsAI SkillsAI GamesAI Short FilmsAI Prompts
Home/AI Skills/agents-best-practices
AI Skills · Codingagents-best-practices

agents-best-practices

Helps you design, audit and refactor an AI agent harness, or draft an MVP blueprint for one.

Ecosystem
Agent Skills
Skill category
Coding
Creator / organization
Denis Shiryaev(DenisSergeevitch)
View work and source ↗

What it is

agents-best-practices is a single, provider-neutral Skill by Denis Shiryaev (DenisSergeevitch). It covers the harness around an agent: how tools are called, how permissions are granted, how the loop runs. You can ask it to design a new harness, generate an MVP blueprint, or audit, refactor and explain one you already have. One principle runs through it: execution and authorization belong to the outer harness, not to the model.

Who it's for:Developers building AI agent products or internal tools. Not useful if you don't write code.

Before you start: Install with npx skills add DenisSergeevitch/agents-best-practices -g, or clone it into your host's folder. No separate service is needed. There is an optional public-board posting feature that shows you the destination and text first and posts only after you approve; the docs forbid attaching private conversations or files. Model costs follow your host. Docs are mainly in English. License: MIT.

Based on the project's public documentation. Not tested by AI Inspiration World.

Used for

  • Agent application development→
  • Software architecture design→

Select one to see other works for the same use.

Source

https://github.com/DenisSergeevitch/agents-best-practices ↗

Repository ↗

How rankings work →
Suggest an update? Contact us
AI Inspiration World logo: aiw. picked by humansAI Inspiration World logo: aiw. picked by humansAI Inspiration World

Check in daily. The newest in AI lives here.

Updated every day · Latest arrival: Oct 31, 2026

Site version: Oct 31, 2026

About·Methodology·Trust & Safety·Contact