DenisSergeevitchDenisSergeevitchCommunityยท1 Agent Skills Included

agents-best-practices

Design, audit, and improve safe AI agent harnesses

Designs production-ready AI agent harnesses with clear loops, tool permissions, memory, and approval gates. Generates MVP blueprints for agents in any domain, from coding and research to finance and support. Audits existing agents for runaway loops, lost context, unsafe tools, and missing evaluations. Replaces vague prompt advice with concrete checklists, budgets, and launch criteria.
npx skills add DenisSergeevitch/agents-best-practices --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install agents-best-practices?โ–ผ

Run `npx skills add DenisSergeevitch/agents-best-practices --all -g -y` in your terminal to install the skill globally for all your projects.

What is an AI agent harness?โ–ผ

A harness is the control layer around an AI model that validates actions, enforces permissions, runs tools, and records results. This skill teaches your agent how to design and audit that layer correctly.

Can it generate an MVP blueprint for my agent idea?โ–ผ

Yes. Describe your domain and the skill produces a concrete MVP blueprint covering the core loop, tool registry, permission matrix, memory, safety gates, and launch criteria.

Does it work with Claude Code and Codex?โ–ผ

Yes. It follows the universal SKILL.md standard and works with Claude Code, Codex, and any compatible agent, across OpenAI, Anthropic, and OpenAI-compatible APIs.

Can it audit my existing AI agent?โ–ผ

Yes. It checks for common failures like missing budgets, lost approval state after compaction, overly broad tools, and absent evaluations, then gives a prioritized fix list.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’