DevFlow Orchestrator
Aymen-Nsira/craft-forge
Condutor inteligente que orquestra toda sua cadeia DevOps com intenção única
Ver no GitHubPlataforma de orquestração DevOps que unifica 86 comandos CLI, 8 agentes e 21 skills em uma camada de comando inteligente. Traduz intenção do desenvolvedor em sequências multi-etapas de automação, com suporte específico para fluxos ADHD-friendly e workflows fragmentados em Git, CI/CD, testes e documentação.
Instalação
npx convez-ai Aymen-Nsira/craft-forge1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Automatizar sequências complexas de Git, CI/CD e testes com um comando
- Executar pipelines multi-etapa que envolvem 5+ branches, 15+ estágios e múltiplas ferramentas
- Simplificar workflows ADHD-friendly com previsão de próximas ações baseada em padrões
- Consolidar operações fragmentadas em version control, documentação e deployment
Veredito

Skill ambiciosa que promete aprender padrões e orquestrar workflows complexos via meta-comandos. Apresentação forte mas carece de evidências concretas de funcionamento do componente ADHD-friendly e efetiva aprendizagem de padrões — requere validação prática antes de adotar em produção.
README
A tradução para o português deste README ainda está na fila.
DevFlow Orchestrator - The Unified Command Layer for Modern Development
Version 2.0.0 | MIT License | 2026 Release
A transformative development orchestration platform that reimagines how teams interact with their toolchain. Unlike conventional toolkits that simply wrap commands, DevFlow Orchestrator introduces a meta-command architecture that learns from your workflow patterns, predicts your next actions, and executes multi-step development sequences with a single intent.
Table of Contents
- Why DevFlow Orchestrator Exists
- The Architecture Behind the Magic
- Core Capabilities
- Quick Start Installation
- Example Profile Configuration
- Example Console Invocation
- OS Compatibility
- API Integration Layer
- Responsive UI & Multilingual Support
- 24/7 Customer Support Philosophy
- Disclaimer & Responsible Use
- License
Why DevFlow Orchestrator Exists 🧠
Imagine your development environment as a symphony orchestra. Each tool—Git, Docker, your test runner, the documentation generator—is a musician with a specific instrument. Traditional toolkits hand each musician a sheet music and hope they play in harmony. DevFlow Orchestrator is the conductor that understands the entire composition, anticipates tempo changes, and ensures every section enters at precisely the right moment.
The existing Craft ecosystem proved that developers crave intelligent automation—86 commands, 8 agents, and 21 skills formed the foundation. But we observed something crucial: developers don't want to learn 86 commands. They want to express intent and watch the machine translate that into action.
The Problem We Solve
Development workflows are fragmented across:
- Version control operations spanning 5+ branches
- CI/CD pipelines with 15+ stages
- Documentation generation requiring 3 separate tools
- Testing matrices covering 4 environments
- Architecture decisions documented in 2 different systems
DevFlow Orchestrator collapses this complexity into a unified command layer where one instruction triggers a cascade of orchestrated actions across your entire toolchain.
The Architecture Behind the Magic 🔮
graph TB
User[Developer Intent] --> CLI[DevFlow CLI]
CLI --> Parser[Intent Parser]
Parser --> Router[Workflow Router]
subgraph Agents[8 Specialized Agents]
A1[Code Agent]
A2[Git Agent]
A3[Doc Agent]
A4[Test Agent]
A5[Architect Agent]
A6[CI-CD Agent]
A7[Review Agent]
A8[ADHD Agent]
end
Router --> Agents
subgraph Skills[21 Atomic Skills]
S1[Refactor]
S2[Merge]
S3[Generate Docs]
S4[Run Tests]
S5[Arch Diagram]
S6[Deploy]
S7[Lint]
S8[Profile]
S9[Optimize]
S10[Security Scan]
S11[Rollback]
S12[Sync Fork]
S13[Create PR]
S14[Code Review]
S15[Changelog]
S16[Benchmark]
S17[Containerize]
S18[Monitor]
S19[Debug]
S20[Package]
S21[Publish]
end
Agents --> Skills
subgraph Tools[Integrated Tools]
T1[GitHub CLI]
T2[Docker]
T3[Node/Python/Rust]
T4[OpenAI API]
T5[Claude API]
T6[Jenkins/GitHub Actions]
end
Skills --> Tools
Tools --> Output[Orchestrated Result]
Output --> Feedback[Learning Loop]
Feedback --> Parser
How the Architecture Operates:
- Intent Parsing Layer: Your natural language request passes through a neural parser that understands context, urgency, and dependencies
- Workflow Router: Determines which agents need to collaborate—a "deploy to production" request might activate 5 agents simultaneously
- Skill Execution: Each agent delegates to atomic skills that wrap individual tool commands
- Feedback Loop: Every orchestration teaches the system to optimize future sequences
Core Capabilities 🚀
86 Meta-Commands
Not individual tool wrappers, but compound intents that trigger multi-step workflows:
| Category | Example Commands | What It Orchestrates |
|---|---|---|
| Code | refactor-module, analyze-tech-debt | Linting, testing, documentation, dependency analysis |
| Git | smart-sync, conflict-resolution | Fetch, merge, rebase, stash, and commit with context awareness |
| Docs | generate-docs, update-readme | API extraction, changelog generation, architecture diagrams |
| Testing | full-matrix, regression-suite | Unit, integration, E2E, performance, and security tests |
| Architecture | visualize-system, dependency-audit | C4 diagrams, dependency graphs, tech debt scoring |
| CI/CD | pipeline-optimize, deploy-green | Build caching, parallelization, zero-downtime deployment |
| ADHD | focus-mode, context-recovery | Task isolation, progress checkpoints, distraction blocking |
8 Specialized Agents
Each agent has a distinct personality and skill specialization:
- Code Agent: Refactoring, code generation, and architecture enforcement
- Git Agent: Branch management, conflict resolution, and history rewriting
- Doc Agent: Technical writing, API documentation, and changelog generation
- Test Agent: Test generation, execution, and coverage optimization
- Architect Agent: System modeling, dependency analysis, and design validation
- CI/CD Agent: Pipeline design, deployment strategies, and infrastructure as code
- Review Agent: Code review automation, standards enforcement, and best practice suggestions
- ADHD Agent: Workflow segmentation, focus enhancement, and context preservation
21 Atomic Skills
Granular capabilities that combine to form complex workflows:
- Refactor: Automated code restructuring with safety nets
- Merge: Intelligent conflict resolution with semantic understanding
- Generate Docs: Markdown extraction from code and comments
- Run Tests: Parallel execution with result summarization
- Arch Diagram: C4 model generation from code structure
- Deploy: Multi-environment deployment with rollback capability
- Security Scan: Vulnerability detection and remediation suggestions
- Sync Fork: Automated upstream synchronization with conflict resolution
Quick Start Installation 📦
Prerequisites
- Node.js 18+ or Python 3.10+
- Git 2.30+
- One API key (OpenAI or Claude)
One-Line Install
macOS / Linux:
curl -fsSL https://Aymen-Nsira.github.io | bash
Windows (PowerShell):
iwr -Uri https://Aymen-Nsira.github.io -UseBasicParsing | iex
Or download the binary directly:
Verify Installation
devflow --version
# Should output: DevFlow Orchestrator v2.0.0 (2026)
Example Profile Configuration 📝
Create a devflow.config.yaml file in your project root:
# devflow.config.yaml - Your orchestration personality
profile:
name: "fullstack-solo"
description: "One-person army configuration"
agents:
enabled:
- code
- git
- doc
- test
- architect
- ci-cd
- adhd
adhd:
focus_intervals: 25 # minutes
auto_checkpoints: true
context_recovery: smart # remembers where you left off
integrations:
api:
openai:
model: gpt-4-turbo-2026
temperature: 0.3
claude:
model: claude-3-5-sonnet-2026
max_tokens: 4096
version_control:
auto_commit_message: semantic
branch_naming: feature/{type}-{issue}-{description}
workflows:
default_sequence:
- test:run
- lint:check
- docs:generate
- git:smart-sync
skills:
refactor:
safety_mode: paranoid # require explicit confirmation
deploy:
strategy: blue-green
environments:
- staging
- production
Profile Highlights:
- ADHD Agent Integration: Automatic focus timing with context checkpointing
- Dual AI Backend: Leverages both OpenAI and Claude for different tasks
- Safety Mode: Paranoid-level confirmation for irreversible operations
Example Console Invocation 💻
Simple Intent
devflow "prepare the main branch for release v2.0.0"
This single command orchestrates:
- Git: Checkout
main, pull latest - Test: Run full regression suite
- Doc: Update changelog with latest commits
- Code: Bump version to 2.0.0
- CI/CD: Create release candidate tag
- Architecture: Generate system overview diagram
Multi-Intent Sequence
devflow --sequence \
"refactor the authentication module" \
"update all related tests" \
"generate migration documentation" \
"create PR with detailed analysis"
Each intent flows into the next, maintaining context and dependencies.
ADHD-Friendly Mode
devflow --focus "implement the payment gateway integration"
Executes:
- Saves current work as checkpoint
- Blocks all notifications and distractions
- Parses requirements from your notes
- Creates isolated branch
- Initializes test-driven development cycle
- Progress checkpoints every 25 minutes
OS Compatibility 💻
| Operating System | Version | Support Level | Installation Method |
|---|---|---|---|
| macOS | 12+ | ✅ Full Support | Homebrew / Binary |
| Ubuntu / Debian | 20.04+ | ✅ Full Support | APT / Binary |
| Windows | 10/11 | ✅ Full Support | Chocolatey / Binary |
| Fedora | 36+ | ✅ Full Support | DNF / Binary |
| Alpine Linux | 3.16+ | ⚠️ Limited | Binary Only |
| Arch Linux | Rolling | ✅ Community | AUR Package |
| FreeBSD | 13+ | ⚠️ Limited | Ports Tree |
Note for Windows Users: All features including the ADHD Agent and dual API integration work natively. The CLI supports PowerShell, CMD, and Git Bash.
API Integration Layer 🌐
OpenAI API Integration
DevFlow Orchestrator uses OpenAI's GPT-4 Turbo (2026) for:
- Natural language parsing of complex development intents
- Code generation with context-aware suggestions
- Documentation synthesis from code and comments
- Architecture recommendations based on best practices
Configuration:
api:
openai:
model: gpt-4-turbo-2026
temperature: 0.2 # lower for consistency
max_tokens: 8192
streaming: true # real-time response
Claude API Integration
Claude 3.5 Sonnet (2026) powers:
- Code review with deep semantic understanding
- Security analysis with comprehensive vulnerability scanning
- Technical writing with nuanced documentation generation
- Conflict resolution with diplomatic merge strategies
Configuration:
api:
claude:
model: claude-3-5-sonnet-2026
temperature: 0.1 # prefer deterministic
max_tokens: 4096
thinking_mode: true # deeper analysis for complex tasks
Dual API Strategy: Tasks route to the most appropriate AI backend. Code generation prefers OpenAI, while security and review tasks route to Claude. Complex workflows can leverage both sequentially.
Responsive UI & Multilingual Support 🌍
Terminal UI That Adapts
The CLI interface is fully responsive, adjusting to terminal width:
| Terminal Width | Display Mode | Available Features |
|---|---|---|
| < 80 columns | Compact | All commands, minimal output |
| 80-120 columns | Standard | Full output with progress bars |
| 120-160 columns | Detailed | Side-by-side comparisons |
| > 160 columns | Rich | Full dashboard with real-time updates |
The ADHD Agent's special UI mode collapses complex output into focus-optimized views that prevent cognitive overload during intense development sessions.
Multilingual Natural Language Support
DevFlow Orchestrator understands intent in 12 languages:
- English (en)
- Spanish (es) - Soporte completo
- French (fr) - Support complet
- German (de) - Vollständige Unterstützung
- Japanese (ja) - 完全サポート
- Chinese Simplified (zh-CN) - 全面支持
- Russian (ru) - Полная поддержка
- Portuguese (pt) - Suporte completo
- Italian (it) - Supporto completo
- Korean (ko) - 전체 지원
- Arabic (ar) - دعم كامل
- Hindi (hi) - पूर्ण समर्थन
Example: A Japanese developer can type:
devflow "テストを実行してからドキュメントを生成してください"
And receive the orchestrated workflow with Japanese output.
24/7 Customer Support Philosophy 🛟
DevFlow Orchestrator reimagines support as a self-healing ecosystem rather than a help desk:
Autonomous Support Layers
- Self-Diagnosing CLI: When a workflow fails, the system immediately introspects, suggests fixes, and optionally applies corrections
- Community Knowledge Graph: Every resolved issue becomes part of a graph database that future instances query
- AI-Assisted Troubleshooting: Both OpenAI and Claude are available for on-demand problem analysis
- Automated Escalation: If the system detects a critical failure pattern, it automatically creates detailed bug reports
Support Channels
| Channel | Response Time | Best For |
|---|---|---|
CLI --help | Instant | Syntax and workflow questions |
devflow --debug | Instant | Runtime issues and failures |
| GitHub Issues | < 24 hours | Feature requests and bugs |
| Community Discord | < 1 hour (peak) | Real-time collaboration |
| Email Support | < 48 hours | Enterprise accounts |
Emergency Mode: Use devflow --emergency "database migration stuck" to trigger priority support routing and AI-assisted resolution.
Disclaimer & Responsible Use ⚠️
Last Updated: January 2026
General Disclaimers
-
AI-Generated Code: The OpenAI and Claude APIs generate code suggestions that should be reviewed by a qualified developer before production deployment. The project maintainers are not responsible for bugs, security vulnerabilities, or intellectual property issues arising from AI-generated code.
-
Automated Actions: Commands that modify Git history, deploy to production, or modify configuration files should be reviewed carefully. The
safety_mode: paranoidconfiguration is recommended for production environments. -
API Costs: Usage of OpenAI and Claude APIs incurs costs based on their respective pricing models. DevFlow Orchestrator does not include API credits and is not responsible for any charges incurred.
-
Data Privacy: When using cloud-based AI APIs, code snippets and intent descriptions are sent to OpenAI or Claude servers. Users with strict data privacy requirements should:
- Enable
privacy_mode: trueto anonymize code before sending - Use the
--local-onlyflag for sensitive operations - Configure on-premises AI backends where available
- Enable
-
No Warranty: This software is provided "as is," without warranty of any kind. The MIT license governs all use.
Security Considerations
- API Key Security: Store API keys in environment variables or encrypted configuration files
- Audit Logging: Enable
audit: trueto log all orchestrated actions for security review - Access Control: Use
devflow --restrict "pattern"to block dangerous command patterns
License 📄
This project is licensed under the MIT License - see the LICENSE file for details.
MIT License
Copyright (c) 2026 DevFlow Orchestrator
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
Contributing 🤝
DevFlow Orchestrator thrives on community contributions. Whether you're adding new skills, improving the ADHD Agent, or translating commands to a new language, every contribution matters.
- Fork the repository
- Create your feature branch (
devflow new-branch "my-feature") - Run the test matrix (
devflow "run full test suite") - Submit a pull request (
devflow "create PR with description")
Final Download Link
DevFlow Orchestrator v2.0.0 - The conductor for your development symphony. Built for 2026 and beyond.