13-chapter book + opinionated .claude/ scaffold + multi-machine sync for senior devs.
Independent third-party product. Not affiliated with or endorsed by Anthropic. "Claude" and "Claude Code" are trademarks of Anthropic.
Claude Code's official docs explain the CLI. They do not explain how to run Claude Code as a daily driver across multiple projects and machines, with rules that compose, hooks that block the commands you regret, skills that actually fire, and memory that does not rot. The free awesome-lists give you 135 agents and zero opinion about which three to keep. The free GitHub templates give you one machine and no sync story. This pack is the opinionated middle.
Inside you get three things. First, a 13-chapter, ~5,700-word book in PDF, EPUB, and plain markdown. It covers the three-layer load order, the rule anti-patterns that silently bloat every session, the hook event model with a working SSH destructive-command blocker, a skill-firing accuracy test harness, the memory graveyard pitfall, an MCP server you can read end to end in 50 lines, the four-reviewer audit pattern, and two operational chapters that become scripts.
Second, an opinionated .claude/ scaffold: CLAUDE.md, settings.json with hooks wired, five rule files, three hooks, three skills, four reviewer agent prompts, and a memory INDEX starter. Third, two scripts no free pack ships — sync-setup.sh wires ~/.claude/ to a private git repo via symlinks with correct machine-local exclusions, and bootstrap-project.sh scaffolds new Node.js, Python, or Go projects with a stack-scoped .claude/ in three minutes.
Install is one command. ./install.sh --global merges into existing ~/.claude/ without clobbering anything (per-file collision detection, --force moves old config to a timestamped backup, --dry-run shows every action). Re-running is a no-op.
Global, project, and per-directory rules. How composition works. When each layer fires. Real examples of scope isolation.
Anti-patterns that silently bloat every session. Quarterly hygiene checklist. How to keep rules DRY.
Pre-tool and post-tool hooks. The workflow for blocking destructive commands on production. Hook reliability across shells.
Why skills fail silently. Testing patterns that catch false positives. Accuracy benchmarking across a corpus.
When memories rot. Deduplication strategies. Preventing the MEMORY.md bankruptcy. Archival patterns.
Model Context Protocol basics. One minimal working example you can read end to end. When to build vs. use existing.
Multi-agent collaboration with reviews. Keeping reviewers in sync. Rollback mechanics for failed batches.
Log inspection techniques. Reproducing bugs in isolation. Symptom-to-root-cause flowchart.
Git symlinks vs. rsync. SSH key management. Idempotency guarantees. Conflict resolution when machines drift.
Scaffolding new projects with bootstrap-project.sh. Stack detection. Templating patterns for Node/Python/Go.
Monitoring Claude Code across machines. Alert patterns. Common failure modes and recovery.
Secrets in rules (what not to do). Hook security model. Restricting .claude/ permissions.
Context window management. Fallback chains across agents. Cache busting strategies. Real production lessons.
Chapter 2 (Rules System: Global, Project, Per-Directory) demonstrates the book's density and opinion quality without revealing the wedge-critical chapters on sync and bootstrap. Read in 5 minutes; see if the architectural reasoning matches your style.
Yes — v1.0 ships with PDF, EPUB, and markdown source. Read in whichever format you prefer.
Buy via Lemon Squeezy (Merchant of Record) — they issue compliant EU VAT invoices automatically. Gumroad does not issue EU VAT invoices; use LS if you need to expense this on a company card.
Yes. The scaffold and scripts work on any POSIX shell (Linux, macOS, WSL). The sync script uses symlinks, which work natively on Linux/macOS and on WSL2.
The architectural patterns (rules layering, hook event model, memory lifecycle) are stable across CLI versions. When Claude Code ships breaking changes, v1.x updates cover them — free for all buyers.
License is single-seat. Team license available on request — contact support at info@shippedstack.com.
No. If you have not yet run your first Claude Code session, start with Anthropic's official quickstart. This pack is for senior developers who already use Claude Code and want the architecture for scaling it across projects and machines.
Free lists give you 100+ links with zero curation. This pack gives you one opinionated architecture: what three agent types to keep, how to wire rules for production safety, how to sync across machines without losing data. You're paying for the opinion, not the links.
Claude Code Power User Guide — €49 one-time, lifetime v1.x updates, 30-day refund.