Samuel¶
Artificial Intelligence Coding Framework
Build smarter, faster, and more scalable software.
What is Samuel?¶
Samuel (Artificial Intelligence Coding Framework) is an opinionated AI development framework designed for professional software teams. It provides guardrails, language guides, and workflows that help AI coding assistants produce consistent, high-quality code.
-
35+ Guardrails
Testable rules for code quality, security, testing, and git practices. Not vague suggestions - specific, enforceable standards.
-
21 Language Guides
All major languages from TypeScript to Zig. Auto-loading based on file extensions - no manual configuration needed.
-
33 Framework Guides
Framework-specific patterns for React, Django, Rails, Spring Boot, and 29 more. Deep integration guidance.
-
24 Workflows
Planning, quality, maintenance, utility, autonomous, and community workflows including PRD creation, code review, security audit, autonomous loops, and more.
-
CLI-Powered
Full-featured CLI with 14 commands: search, info, add, remove, diff, config, skill, auto, sync, and more. Discover components before installing.
-
Cross-Tool Compatible
Works with Claude Code, Cursor, Codex, Copilot, and any AI assistant that reads AGENTS.md.
Quick Start¶
# 1. Install the CLI
brew tap ar4mirez/tap && brew install samuel
# Or: curl -sSL https://raw.githubusercontent.com/ar4mirez/samuel/main/install.sh | sh
# 2. Initialize your project
cd your-project
samuel init
# 3. Discover and add components
samuel search react # Find components
samuel info fw react --preview # Preview before installing
samuel add framework react # Install what you need
# 4. Verify installation
samuel doctor
The system works immediately:
- AI loads CLAUDE.md automatically (500 lines of guardrails + operations)
- Language guides auto-load based on file extensions
- 14 CLI commands for component discovery and management
- Progressive - starts minimal, grows with your project
The 3 Modes¶
AI auto-detects which mode to use based on your request:
For: Bug fixes, small features (<5 files)
- Direct implementation
- Quick validation
- One commit
For: New components, API endpoints (5-10 files)
- Break into 3-5 subtasks
- Implement sequentially
- Integration testing
Key Guardrails¶
Security (CRITICAL)
- All inputs validated before processing
- Parameterized queries only (no string concatenation)
- No secrets in code
- Dependencies checked for vulnerabilities + licenses
Code Quality
- Functions ≤50 lines
- Files ≤300 lines
- Cyclomatic complexity ≤10
- All exports have types/docs
Testing (CRITICAL)
-
80% coverage for business logic
-
60% overall coverage
- Tests for all public APIs
- Regression tests for bugs
The 4D Methodology¶
graph LR
A[DECONSTRUCT] --> B[DIAGNOSE]
B --> C[DEVELOP]
C --> D[DELIVER]
A -.- A1[Break down the task]
B -.- B1[Identify risks & dependencies]
C -.- C1[Implement with tests]
D -.- D1[Validate & commit] What's New in v2.0.0¶
- Autonomous AI Coding Loop - Ralph Wiggum methodology for unattended AI-driven development (
samuel auto) - Native
.claude/Integration - Migrated from.agent/to.claude/native directory - Agent Skills Standard - Portable skill modules across 25+ AI tools (
samuel skill) - Per-folder CLAUDE.md - Hierarchical instructions with folder-level overrides
System Stats¶
| Metric | Value |
|---|---|
| Version | 2.0.0 |
| Status | Production Ready |
| AGENTS.md | Compatible |
| Total Files | 70+ markdown files |
| CLAUDE.md | ~500 lines |
| CLI Commands | 14 (init, search, info, add, remove, list, config, diff, update, doctor, version, skill, auto, sync) |
| Language Guides | 21 (all major programming languages) |
| Framework Guides | 33 (across 11 language families) |
| Workflows | 24 (planning, quality, maintenance, utility, autonomous, community) |
| Guardrails | 35+ testable rules |
Philosophy¶
Small, validated changes. Quality enforced. Documentation grows organically.
This framework embraces:
- Atomic commits - One logical change per commit
- Progressive disclosure - Start minimal, grow with your project
- Cross-tool compatibility - Works with any AI coding assistant
- Testable standards - Not vague suggestions, but enforceable rules
-
Documentation
Comprehensive guides for every feature.
-
GitHub
Star, fork, or contribute.
-
Issues
Report bugs or request features.
-
Discussions
Ask questions and share ideas.