DenisSergeevitch/agents-best-practices

Provider-neutral Agent Skill for Codex, Claude Code, and agentic harness design.

What it solves

This project provides a structured framework and set of best practices for building "agentic harnesses"—the runtime environment that surrounds an AI model to make it safe, reliable, and production-ready. It addresses the tension between giving an agent autonomy to perform useful work and preventing the model from becoming an unaudited, unsafe operator in real-world systems.

How it works

Instead of relying on prompt-based safety, the project defines a rigorous control plane where the model only proposes actions, while the application code (the harness) handles validation, authorization, and execution. It provides a comprehensive library of reference documents covering:

  • The Agentic Loop: Managing retries, budgets (time, cost, tokens), and termination rules.
  • Tool Governance: Using narrow, typed tools and risk-based permission levels (e.g., separating reads, drafts, and destructive writes).
  • Context Management: Strategies for retrieval, memory, and auto-compaction that preserve active state.
  • Advanced Architectures: Blueprints for MVP agents, recursive harnesses, and speculative tool execution.
  • Environment Adaptation: Methods for agents to safely discover and bind to tools in unfamiliar environments.

Who it’s for

Developers and architects building AI agents for various domains—including coding, research, support, operations, sales, finance, and healthcare—who need to move beyond simple prototypes to production-safe systems.

Highlights

  • Provider-Neutral: Works across different LLM providers (OpenAI, Anthropic, etc.).
  • Risk-Based Permissions: Enforces a clear distinction between autonomous reads and approval-gated writes.
  • Harness-Centric Philosophy: Shifts the burden of safety and reliability from the prompt to the runtime code.
  • Comprehensive Blueprints: Includes specific guides for MVP design, auditing existing harnesses, and implementing long-running goals.
  • Agent Skill Integration: Packaged as a portable "Agent Skill" compatible with Codex and Claude Code.

Related

  • Project
    HoangNguyen0403/agent-skills-standard

    Agent Skills Standard (AGS) is an NPM‑distributed CLI that lets teams publish their coding‑standards, security rules, and SDLC workflow templates as tiny markdown “skills”. The CLI (`ags init` / `ags sync`) creates a `.skillsrc` config, downloads language‑specific skill packs (280 skills across 20+ stacks), and writes a hierarchical router (`AGENTS.md → _INDEX.md → SKILL.md`). AI coding assistants (Cursor, Claude Code, Copilot, Gemini, etc.) read this router and load only the relevant rules for the file they are editing, saving ~85 % of prompt tokens. An optional MCP server can enforce hooks, block secret‑leak edits, and provide runtime audit logs. The project is a genuine AI‑coding‑assistant utility, not a generic tutorial or list.

  • Project
    mohitagw15856/pm-claude-skills

    PM Skills is an open‑source library of 1 169 markdown “Agent Skills” that give Claude, ChatGPT, Gemini, etc. concrete, senior‑level procedures for professional and life‑admin tasks (e.g., lease decoding, PRD writing, salary‑negotiation simulation). Skills are plain markdown files containing framework, template, checks, and anti‑patterns. They can be installed with a one‑line `npx` command, used natively in Claude Code, or pasted into prompts for other agents. The repo also provides npm/PyPI packages, Docker image, and integrations (Telegram, Slack, Raycast, Obsidian, n8n). A web playground lets anyone run a skill in the browser without installing anything.

  • Project
    shareAI-lab/learn-claude-code

    A 17‑lesson tutorial that teaches you how to build the surrounding “harness” code for an LLM‑based coding agent (Claude). It walks through adding tools, permissions, context management, memory, task scheduling, multi‑agent coordination, and goal‑driven termination, all around a simple agent‑loop that lets the model decide when to call tools.

  • Project
    addyosmani/agent-skills

    Agent Skills is a library of 25 Markdown‑based engineering workflows (spec, plan, build, test, review, ship, etc.) that can be installed into many AI coding agents via a simple CLI. It provides slash commands, auto‑build mode, specialist personas, and reference checklists to make AI‑generated code follow senior‑engineer best practices across the entire development lifecycle.

  • Project
    MoizIbnYousaf/ai-agent-skills

    AI Agent Skills is an npm‑based CLI/TUI that ships a curated library of ~115 AI‑agent “skills” (prompt‑code bundles) organized into shelves (frontend, backend, mobile, workflow, agent‑engineering, marketing). It lets you browse, search, install, and manage both local “house‑copy” skills and upstream metadata‑only skills. You can also bootstrap a personal library with `init-library`, add or catalog new skills, vendor copies, sync updates, and generate documentation. Works with any Agent‑Skills‑compatible agent (Claude, Cursor, Gemini, etc.) and supports multiple install targets. Ideal for developers or teams who want a tidy, searchable toolbox of reusable agent capabilities.