claude-code-bootstrap

The Bootstrap Protocol: AI Infrastructure-as-Code A reusable framework to instantly configure any repository for high-performance AI-assisted development.

The Core Concept

Most AI configuration is static and brittle. The Bootstrap Protocol introduces a Generator Pattern: you maintain a master library of generic templates, and a single /claude-code-bootstrap command hydrates them by scanning your project's stack. This transforms Claude AI from a general-purpose chatbot into a specialized Project Architect that understands your specific paths, patterns, and dependencies.

Quick Start

# 1. In your project root — sets up the .claude/ folder
npx claude-code-bootstrap

# 2. Open Claude Code in this directory, then run:
/claude-code-bootstrap

That's it. 60+ commands tailored to your stack are ready to use.

The Three-Phase /claude-code-bootstrap Logic

Phase 1 — Discovery (The Scan)

The AI performs a rapid, multi-layered environment check:

• Stack Check — Identifies languages and frameworks (package.json, go.mod, requirements.txt)
• Path Check — Maps where components, routes, and tests live
• Env Check — Reads .env.example to identify required service integrations

Phase 2 — Hydration (The Fill)

Discovered data is mapped to template placeholders:

{{BE_TEST_COMMAND}}      →  npm run test:unit
{{FE_PATH_COMPONENTS}}   →  src/app/components
{{GRAPH_IMPORTANCE_LOGIC}} →  (maps to high-traffic files identified by Graphify)

Phase 3 — Deployment (The Output)

The AI generates three foundational artifacts:

• CLAUDE.md — The Session Warm-up. Stack, paths, and commands baked-in so every new session starts with zero re-discovery.
• .claudeignore — The Token Shield. Automatically excludes node_modules, dist, .venv, and secrets to optimize the context window.
• commands/ — A pruned library of executable instructions. No Docker files? No cloud commands generated.

After hydration, .claude/templates/ is automatically deleted — only your project-specific commands/ remains.

System Structure

.claude/
├── templates/              # SOURCE: Pristine generic templates (temporary)
│   ├── backend/            # FastAPI, Express, Django patterns
│   ├── frontend/           # React, Next.js, Vue patterns
│   ├── cloud/              # Docker, K8s, CI/CD patterns
│   ├── chains/             # Multi-step workflow playbooks
│   ├── general/            # Cross-stack logic
│   └── review/             # Quality gates & audits
├── commands/               # LIVE: Hydrated project-specific commands
│   ├── claude-code-bootstrap.md   # The Architect — run this first
│   └── ...                 # Pruned, project-ready commands
├── settings.json           # Global Claude settings
└── settings.local.json     # THE GUARDRAILS: Safety overrides

Advanced Features

Knowledge Graph Integration

Run /graphify before /claude-code-bootstrap for deeper intelligence:

/graphify              # generates graphify-out/GRAPH_REPORT.md
/claude-code-bootstrap # reads graph output to inject importance scores

Commands like /general:research_codebase read this cached snapshot to understand "Core Nodes" and file relationships without re-crawling, saving thousands of tokens.

Intelligent Model Routing

| Model | Commands | |-------|----------| | Haiku (Fast/Cheap) | /commit, /describe_pr, /create_handoff | | Sonnet (Balanced) | Most implementation and testing tasks | | Opus (Deep Reasoning) | /create_plan, /smart-debug, /research_codebase |

Workflow Chains

End-to-end playbooks instead of micro-managing:

| Chain | Steps | |-------|-------| | /chains:feature-workflow | Onboard → Research → Plan → Implement → Test → PR | | /chains:bugfix-workflow | Debug → Smart-Debug → Fix → Test → Validate → PR | | /chains:migration-workflow | Research → Migrate → DB Migrate → Test → Deploy | | /chains:security-workflow | Backend → Frontend → Cloud → Compliance → Deps → Deploy |

Safety & Governance (settings.local.json)

Policy-Based Access Control lets the AI work autonomously without risking your repository:

| Policy | Operations | Reason | |--------|-----------|--------| | Hard Deny | git force-push, rebase, merge, reset --hard | Preserve human Source of Truth | | Ask | rm -rf, destructive deletes | Require manual confirmation | | Allow | Read, Edit, Write, Grep, Glob | Friction-free for maximum velocity |

Template Library (63 templates)

| Category | Count | Examples | |----------|-------|---------| | backend/ | 11 | api-scaffold, db-migrate, security-scan, test-harness | | frontend/ | 11 | component-scaffold, accessibility-audit, performance-audit | | cloud/ | 8 | docker-optimize, k8s-manifest, monitor-setup | | general/ | 21 | create_plan, smart-debug, research_codebase, issue | | review/ | 7 | pr-enhance, deps-audit, compliance-check | | chains/ | 5 | feature-workflow, bugfix-workflow, security-workflow |

Re-initialization

If you add Docker, switch frameworks, or onboard a new service:

npx claude-code-bootstrap --templates-only   # restore templates
/claude-code-bootstrap                        # re-hydrate for new stack

Options

npx claude-code-bootstrap                    # full setup, skip existing files
npx claude-code-bootstrap --force            # overwrite all files
npx claude-code-bootstrap --dry-run          # preview what would be copied
npx claude-code-bootstrap --settings-only    # copy only settings files
npx claude-code-bootstrap --templates-only   # restore templates to re-run /claude-code-bootstrap

Developer Benefits

| Feature | Benefit | |---------|---------| | Portability | Drop into 100 repos — adapts to each stack instantly | | Idempotency | Re-run when stack changes — preserves custom notes | | Safety | Destructive Git ops blocked by default | | Consistency | Same commands (/test, /create_plan) regardless of language |

License

MIT © Truedax