OthmanAdi

Agent Blueprint — AI skill for Claude Code

AI community

Build production-grade AI agents from scratch.

How to install Agent Blueprint

This entry records only its repository, not the path inside it, so there is no exact command to give. Open OthmanAdi/agent-blueprint and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Blueprint does

Build production-grade AI agents from scratch. Patterns, templates, and references for workflow builders and app embedders.

Alternatives in AI

Claude Code Configs — A comprehensive collection of production-grade Claude Code configurations, specialized agents, and automation 624 ★
Unofficial Claude Code Prompt Playbook — Unofficial playbook for production-grade LLM system prompt architecture, derived from local analysis of Claude 258 ★
Claudecode Harness — Production-grade orchestration harness for Claude Code — Context discipline, subagents, and multi-model consen 223 ★

README

agent-blueprint

Build production-grade AI agents from scratch. Complete patterns, templates, and reference implementations for any language or framework.

What This Is

A skill for AI coding agents (Claude Code, Cursor, etc.) that guides you through building complete agentic systems — tool registries, permission systems, context management, streaming APIs, human-in-the-loop workflows, and HTTP serving layers.

**Two audiences:**

Workflow builders — planner + executor pipelines, HITL checkpoints, multi-agent orchestration
Application embedders — HTTP/SSE APIs, multi-user session management, React frontend integration

Install

npx skills add OthmanAdi/agent-blueprint -g

Works with Claude Code, Cursor, and any agent that supports the skills protocol. Then just describe what you want to build — the skill activates automatically.

What's Inside

agent-blueprint/
  SKILL.md                    # Main skill — loaded by the agent
  references/                 # Deep-dive docs (load on demand)
    agent-loop.md             # Streaming agent loop (Python)
    typescript-agent-loop.md  # Streaming agent loop (TypeScript/Bun)
    tool-system.md            # Tool registry & execution pipeline
    permission-system.md      # 7-layer permission pipeline
    context-management.md     # Context window lifecycle
    system-prompts.md         # Prompt caching patterns
    mcp.md                    # MCP server integration
    memory.md                 # Session persistence & long-term memory
    rust-agent.md             # Rust implementation (tokio + reqwest)
    production-principles.md  # 85% compounding problem, Manus principles
    planner-executor.md       # Three-agent pattern (planner/executor/verifier)
    human-in-the-loop.md      # HITL approval matrix & tiered delegation
    serving.md                # FastAPI + SSE, multi-user sessions