Clone to Claude
Last commit on June 29, 2026·Created on May 15, 2026
DenisSergeevitch/agents-best-practices
“Blueprints the control plane for agents that run anywhere.”
Combined rank
#112
across all skills
In Prompting
#4
category rank
Stars
2.2k
+0.9% in last 7d
Forks
196
+2.6% in last 7d
Watchers
11
0.0% in last 7d
Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of real attention, adoption, and active maintenance.
Description
This skill is a complete architecture guide for building agentic harnesses that are provider-neutral, safe, and ready for evaluation. It covers MVP blueprint generation, tool registries, permission matrices, context compaction, and non-negotiable principles for reliability. The reference map helps platform teams, power users, and governance teams design harnesses that are rigorous without being rigid.
WHO IT'S FOR
Agent Infrastructure Engineers
designing provider-neutral agent harnesses
AI Product Managers
scoping MVP agent blueprints for new domains
Platform / DevEx Teams
standardizing agent harness patterns across teams
AI Safety / Governance Teams
auditing agent harnesses for reliability and safety
30 Days of GitHub Stars
+100 stars↑ 4.9% monthly change
Jul 7Jul 14Jul 21Jul 28Aug 5
Compatibility & Install1 Skill · 2 Compatible agents · 4 categories · 3 install paths
Compatible agents
❯Claude Code❯Codex
Categories
PromptingAgent OrchestrationSoftware EngineeringDeveloper Tooling
Install viaRepository instructions Repository instructions
Repository instructions
Clone to Codex
NPX Skills Add
Also in this category4 of 13