The Claude Code Playbook

Stop prompting. Start engineering.

Built by Force Information Systems · A Harris Computer Company · Part of Constellation Software

Battle-tested patterns from 900+ sessions across production TypeScript projects.

What’s Inside

Category Count Description
Skills 29 Production-ready custom /commands you can drop into any project
Templates 11 Stack-specific CLAUDE.md files for TypeScript, React, Node, Python, Go, Rust, and more
Hooks 9 Pre-commit and pre-push guard scripts that catch errors before they reach your commits
Docs 30+ Guides, patterns, anti-patterns, troubleshooting, and reference material
Examples 5 Annotated real-world sessions showing workflows in action
Onboarding 6 Step-by-step guides from installation to advanced usage

The Core Loop

Every successful Claude Code session follows the same rhythm:

flowchart LR
    A([Request]) --> B[Implement]
    B --> C{Verify}
    C -- pass --> D([Close])
    C -- fail --> B
    D --> A
    classDef start fill:#1a2b4a,stroke:#0f1a2e,color:#fff,rx:10,ry:10;
    classDef action fill:#f4f7fb,stroke:#1a2b4a,color:#1a2b4a;
    classDef gate fill:#ffd966,stroke:#a07800,color:#4d3800;
    class A,D start;
    class B action;
    class C gate;

The cardinal rule: Each step has a clear boundary. Don’t blur them. Plan in one session, execute in another. Verify with real tests, not code inspection.

The Claude Code Harness

The pieces that sit between you and the raw model — and how they work together:

flowchart TB
    You([You])
    CC[Claude Code CLI]
    Model[Claude Opus 4.7]
    CLAUDE[CLAUDE.md<br/>advisory context]
    Skills["/slash-commands<br/>skills ecosystem"]
    Hooks[Hooks<br/>deterministic enforcement]
    MCP[MCP servers<br/>tools & data]
    Sub[Sub-agents<br/>fresh contexts]

    You --> CC
    CC <--> Model
    CLAUDE -.-> CC
    Skills -.-> CC
    Hooks --> CC
    MCP <--> CC
    CC --> Sub
    Sub --> CC

    classDef user fill:#1a2b4a,stroke:#0f1a2e,color:#fff,rx:10,ry:10;
    classDef tool fill:#f4f7fb,stroke:#1a2b4a,color:#1a2b4a;
    classDef config fill:#e8f1fa,stroke:#4a6fa5,color:#1a2b4a,stroke-dasharray:4 3;
    classDef enforce fill:#ffd966,stroke:#a07800,color:#4d3800;
    class You user;
    class CC,Model,Sub tool;
    class CLAUDE,Skills config;
    class Hooks,MCP enforce;

Read: solid arrows = runtime data flow, dashed = advisory context loaded into prompts, MCP/Hooks = enforcement boundaries you control.

Quick Start

  1. New to Claude Code? Start with the Getting Started guide
  2. Want the full picture? Read the 667-line Complete Guide
  3. Need a quick reference? Grab the Cheat Sheet
  4. Setting up a project? Pick a CLAUDE.md template for your stack
  5. Want custom commands? Browse the Skills Ecosystem

🔖 April 2026 Briefing

The shareable 20-minute summary of this month’s AI development shifts. Covers Opus 4.7, cost optimisation, multi-model orchestration, regulated-AI compliance, and a 10-item 30-day action list for engineering leaders.

→ Read the April 2026 Briefing

Backed by 57 deep-read article pages in News & Research across 9 categories — Models & Vendors, Cost & Performance, Multi-Model Orchestration, Prompting & Discipline, Knowledge Management, Local Inference, Claude Code Features & Updates, Industry Signals, and Adjacent & Off-Topic.

The playbook now covers the Opus 4.7 release and the related ecosystem shifts:

  • Opus 4.7 Reference — the five behavioural patterns, xhigh default, /ultrareview, task budgets, 1M context, Cyber Verification Program, and the migration checklist
  • Cost & Observability — the caveman output-compression plugin (75% reduction) and Rezvani’s OpenTelemetry monitoring stack (Docker Compose + 8 metrics)
  • Regulated AI — SR 11-7 and EU AI Act Article 12 implications for multi-model AI pipelines (Critique vs Model Council), with the three vendor questions to ask before signing
  • Multi-Model Orchestration — the official OpenAI Codex plugin for Claude Code, the CLI-vs-MCP 70/30 decision framework, Anthropic Managed Agents, and the loop-breaker heuristic
  • Prompt Discipline — Cialdini persuasion principles applied to CLAUDE.md, rationalisation tables, 13 red flags, and pressure-testing as TDD-for-prompts
  • Knowledge & Context — Karpathy’s LLM Wiki pattern and the 5-project ecosystem (Waykee Cortex, Sage-Wiki, Thinking-MCP, ELF, qmd)
  • Local Models — Gemma 4 (Apache 2.0) as a local alternative in Codex CLI via llama.cpp or Ollama
  • BMad Autonomous Development — the /bad coordinator for overnight sprint execution with git-worktree-per-story isolation

Why This Exists

Most Claude Code guides tell you how to install it. This one tells you how to use it well.

After months of daily production use — debugging at 2am, shipping features across 30+ file changes, managing fleets of sub-agents, and learning the hard way what breaks — we distilled everything into this playbook.

Built by Force Information Systems · Harris Computer · Constellation Software. Licensed under MIT.