https://github.com/doncheli/don-cheli-sdd

Don Cheli — SDD Framework. The most comprehensive Specification-Driven Development framework for AI agents. 71+ commands, 42 skills, 15 reasoning models. TDD mandatory, OWASP audit, Anthropic Skills 2.0 compatible. Works with Claude Code, Gemini/Antigravity, Cursor, Codex. ES/EN/PT.
https://github.com/doncheli/don-cheli-sdd
agentic-coding ai-agents ai-coding ai-development anthropic-skills claude claude-code developer-productivity developer-tools framework latin-america llm-tools multilingual open-source owasp prompt-engineering sdd spanish specification-driven-development tdd
Last synced: 1 day ago
JSON representation
Host: GitHub
URL: https://github.com/doncheli/don-cheli-sdd
Owner: doncheli
License: apache-2.0
Created: 2026-03-21T23:18:51.000Z (18 days ago)
Default Branch: main
Last Pushed: 2026-03-31T17:49:55.000Z (8 days ago)
Last Synced: 2026-03-31T18:14:55.602Z (8 days ago)
Topics: agentic-coding, ai-agents, ai-coding, ai-development, anthropic-skills, claude, claude-code, developer-productivity, developer-tools, framework, latin-america, llm-tools, multilingual, open-source, owasp, prompt-engineering, sdd, spanish, specification-driven-development, tdd
Language: Shell
Homepage: https://doncheli.tv
Size: 2.08 MB
Stars: 19
Watchers: 5
Forks: 6
Open Issues: 0
Metadata Files:
- Readme: README.md
- Security: SECURITY.md
- Notice: NOTICE
- Agents: AGENTS.md
Awesome Lists containing this project

README

          > 🌐 Read this in other languages: [English](README.md) | [Español](README.es.md) | [Português](README.pt.md)



  
Don Cheli — SDD Framework

  

    Stop guessing. Start engineering.


    _{Vibe coding is the spark; SDD is the engine. Transition from AI-assisted chaos to professional software delivery.}

  

  

    The most comprehensive Specification-Driven Development framework on the market.


    Open source. Multilingual (ES/EN/PT). For Claude Code and other AI agents.

  

  

    

    

    

    

    

    

    

    

    


    

    

    

    

    

    

  


---

## Demo

```bash

# Without Don Cheli:

"Claude, build me a users API"

# → Code without tests → broken in production → "What did we decide yesterday?"

# With Don Cheli (one command):

/dc:start "Users API with JWT authentication"

# → Gherkin Spec → Tests first → Code → Review → Done with evidence

```

> **How does it look in action?** Type `/dc:start` and Don Cheli auto-detects complexity,

> generates the Gherkin spec, proposes the technical blueprint,

> breaks down into TDD tasks and executes. No vibe coding. With evidence.

---

## The Problem

You start a project with AI. The first 2 hours go well. Then:

- **Context rot** — Claude forgets your architecture decisions

- **Silent stubs** — It says "I implemented the service" but the code says `// TODO`

- **No verification** — Does it work? I don't know. Tests? No. Can I deploy? Hopefully

That is **vibe coding**. And it is the enemy of quality software.

---

## Before vs After

| Aspect | ❌ Without Don Cheli | ✅ With Don Cheli |

|--------|---------------------|------------------|

| **Requirements** | "Build me a login" | Gherkin spec with 8 verifiable scenarios |

| **Architecture** | AI invents on the fly | Technical blueprint + ratified DBML |

| **Tests** | "Maybe... someday..." | Mandatory TDD: RED → GREEN → REFACTOR |

| **Quality** | "I think it works" | 6 Quality Gates + 85% coverage |

| **Context** | Lost every session | Full persistence in `.dc/` files |

| **Stubs** | Ship to production | Automatic ghost stub detection |

---

## Installation

**3 steps. 2 minutes. Free.**

### Via npm (recommended)

```bash

# 1. Install globally

npm install -g don-cheli-sdd

# 2. Run the interactive installer

don-cheli install --global

# 3. In your project, open your AI agent and type:

/dc:init

```

### VS Code Extension

Search **"Don Cheli SDD"** in VS Code Extensions, or:

```bash

code --install-extension doncheli.don-cheli-sdd

```

Gives you: sidebar with project status, quality gates, commands browser and metrics dashboard.

### Via git clone

```bash

git clone https://github.com/doncheli/don-cheli-sdd.git

cd don-cheli-sdd && bash scripts/instalar.sh

```

Via npx (no install)

```bash

npx don-cheli-sdd install --global --lang en

```

Remote install (one liner)

```bash

curl -fsSL https://raw.githubusercontent.com/doncheli/don-cheli-sdd/main/scripts/instalar.sh | bash -s -- --global --lang en

```

Silent install (CI/CD)

```bash

bash scripts/instalar.sh \

  --tools claude,cursor \

  --profile phantom \

  --global --lang en

```

Flags: `--tools`, `--profile`, `--skills`, `--comandos`, `--dry-run`, `--global`, `--lang`

The interactive installer guides you step by step:

```

┌──────────────────────────────────────┐

│  Don Cheli SDD — Setup               │

└──────────────────────────────────────┘

Step 1: 🌍 Language → Español, English, Português

Step 2: 🔧 Tool     → Claude Code, Cursor, Antigravity, Codex, Warp, Amp...

Step 3: 👤 Profile  → 6 preconfigured archetypes

Step 4: ✅ Confirm  → Summary of everything selected

```

**Requirements:** Git + an AI agent (Claude Code, Cursor, etc.)

---

## How It Works

**6 phases. From idea to verified code.**

```mermaid

flowchart LR

    P["📋 PRD"] -.->|optional| B

    A["💡 Idea"] --> B["📄 Specify"]

    B --> C["🔍 Clarify"]

    C --> D["🏗 Plan"]

    D --> E["📋 Break Down"]

    E --> F["⚡ Implement"]

    F --> G["✅ Review"]

    style P fill:#2d3436,color:#dfe6e9,stroke:#6c5ce7,stroke-dasharray: 5 5

    style B fill:#6c5ce7,color:#fff

    style C fill:#0984e3,color:#fff

    style D fill:#00b894,color:#fff

    style E fill:#fdcb6e,color:#000

    style F fill:#e17055,color:#fff

    style G fill:#fd79a8,color:#fff

```

> **PRD (optional):** If you have a Figma design, product brief or user research, run `/dc:prd` first. The PRD auto-feeds into Specify with user stories, priorities and risks.

| # | Phase | Command | What it does |

|---|-------|---------|-------------|

| 1 | **Specify** | `/dc:specify` | Turns your idea into a Gherkin specification with test scenarios, priorities and DBML schema |

| 2 | **Clarify** | `/dc:clarify` | A virtual QA detects ambiguities and contradictions before coding |

| 3 | **Plan** | `/dc:tech-plan` | Technical blueprint with architecture, API contracts and final schema |

| 4 | **Break Down** | `/dc:breakdown` | Splits the plan into concrete tasks with execution order and parallelism |

| 5 | **Implement** | `/dc:implement` | Executes with strict TDD: test first, then code, then improve |

| 6 | **Review** | `/dc:review` | Automatic peer review across 7 dimensions: functionality, tests, performance, architecture, security, maintainability, docs |

Each phase has **quality gates**. You don't advance without passing. **No shortcuts.**

---

## Adapts to Your Project

Not everything needs all 6 phases. Don Cheli auto-detects complexity:

| Level | Name | When | Phases |

|-------|------|------|--------|

| **0** | Atomic | 1 file, < 30 min | Implement → Verify |

| **P** | PoC | Validate feasibility (2-4h) | Hypothesis → Build → Evaluate → Verdict |

| **1** | Micro | 1-3 files | Specify (light) → Implement → Review |

| **2** | Standard | Multiple files, 1-3 days | All 6 phases |

| **3** | Complex | Multi-module, 1-2 weeks | 6 phases + pseudocode |

| **4** | Product | New system, 2+ weeks | 6 phases + constitution + proposal |

```bash

/dc:start Implement JWT authentication

# → ▶ Level detected: 2 — Standard

# → ▶ Phases: Specify → Clarify → Plan → Break Down → Implement → Review

```

---

## The 3 Iron Laws

Non-negotiable. Always enforced. No exceptions.

| Law | Principle | In practice |

|-----|-----------|-------------|

| **I. TDD** | All code requires tests | `RED` → `GREEN` → `REFACTOR`, no exceptions |

| **II. Debugging** | Root cause first | Reproduce → Isolate → Understand → Fix → Verify |

| **III. Verification** | Evidence before assertions | ✅ "Tests pass" > ❌ "I think it works" |

---

## Why Don Cheli

BMAD
_{41K ⭐}GSD
_{38K ⭐}spec-kit
_{40K ⭐}Don Cheli

Commands~20~80~1085+

Skills~15~15~642+

Reasoning models———15

Automatic estimates———4 models

Formal quality gates—146

Mandatory TDD———Iron Law

PoC Mode———✅

OWASP Audit———✅

Stack Migration———✅

Stub Detection—✅—✅

UI/API Contracts—✅—✅

Multilingual (ES/EN/PT)———✅

Anthropic Skills 2.0———✅

Worktree Isolation———✅

Crash Recovery———✅

Cost Tracking———✅

Loop Detection———✅

Skills Marketplace———✅

Adversarial Multi-role Debate———✅

CI/CD GitHub Action———✅

Custom Quality Gates (plugins)———✅

Telemetry Dashboard———✅

VS Code Extension———✅

Drift Detection (async watcher)———✅

Reasoning Time Travel———✅

Pre-Flight Cost Simulation———✅

20 things only Don Cheli has

1. **15 reasoning models** — Pre-mortem, 5 Whys, Pareto, RLM

2. **4 estimation models** — Function Points, AI Planning Poker, COCOMO, Historical

3. **PoC Mode** — Validate ideas with timebox and success criteria before committing

4. **Blueprint Distillation** — Extract specs from existing code (behavior reverse engineering)

5. **CodeRAG** — Index reference repos and retrieve relevant patterns

6. **OWASP Audit** — Static security scanning integrated into the pipeline

7. **Stack Migration** — Vue→React, JS→TS with wave plan and equivalences

8. **API Contracts** — REST/GraphQL with retries, circuit breaker, idempotency

9. **SOLID Refactoring** — Checklist, metrics, structured design patterns

10. **Living Documentation** — ADRs, auto-generated OpenAPI, Mermaid diagrams

11. **Captures & Triage** — Capture ideas without stopping, auto-classification in 5 categories

12. **Auto-generated UAT** — Human-executable acceptance scripts after each feature

13. **Doctor** — Diagnosis and auto-repair of git, framework and environment

14. **Skill Creator** — Iterative meta-skill for creating skills automatically

15. **Skills Marketplace** — Install skills from Anthropic, community, or create your own

16. **Project Constitution** — 8 immutable principles validated at every quality gate

17. **Formal Pseudocode (SPARC)** — Technology-agnostic logical reasoning

18. **Multi-layer Validation** — 8 checks (leakage, measurability, completeness, constitution)

19. **Adversarial Debate** — PM vs Architect vs QA with mandatory objection

20. **Scale-adaptive Planning** — Process adjusts by complexity (N0 to N4)

---

## Profiles

6 preconfigured archetypes. Each with optimized skills, commands and reasoning models:

| Profile | Role | Best for | Reasoning |

|---------|------|----------|-----------|

| 👻 **Phantom Coder** | Full-stack | Full pipeline, TDD, quality gates, deploy | First Principles, Pre-mortem, 5 Whys |

| 💀 **Reaper Sec** | Security | OWASP, audits, pentest, offensive/defensive security | Pre-mortem, Inversion, First Principles |

| 🏗 **System Architect** | Architecture | Blueprints, SOLID, APIs, migrations, system design | First Principles, Second Order, Map-Territory |

| ⚡ **Speedrunner** | MVP/Startup | Quick PoCs, agile estimates, ship first | Pre-mortem, Pareto, Opportunity Cost |

| 🔮 **The Oracle** | Reasoning | 15 mental models, deep analysis, hard decisions | All 15 models |

| 🥷 **Dev Dojo** | Learning | Living docs, ADRs, reflections, grow while building | First Principles, 5 Whys, Second Order |

---

## Commands (85+)

Top 20 most used. [Full list in the web docs →](https://doncheli.tv/comousar.html)

### Main pipeline

| Command | What it does |

|---------|-------------|

| `/dc:start` | Start a task auto-detecting complexity (Level 0-4) |

| `/dc:specify` | Convert your idea into Gherkin spec with test scenarios |

| `/dc:clarify` | Find ambiguities and resolve them before coding |

| `/dc:tech-plan` | Generate technical blueprint with architecture and contracts |

| `/dc:breakdown` | Split plan into concrete tasks with execution order |

| `/dc:implement` | Execute tasks with TDD: RED → GREEN → REFACTOR |

| `/dc:review` | Automatic peer review in 7 dimensions |

### Analysis and decisions

| Command | What it does |

|---------|-------------|

| `/dc:explore` | Explore the codebase before proposing changes |

| `/dc:estimate` | Estimates with 4 models (Function Points, COCOMO, Planning Poker, Historical) |

| `/dc:roundtable` | Multi-perspective discussion: CPO, UX, Business |

| `/dc:tech-panel` | Expert panel: Tech Lead, Backend, Frontend, Architect |

| `/dc:security-audit` | Static OWASP Top 10 audit |

| `/dc:poc` | Proof of Concept with timebox and clear criteria |

### Session and context

| Command | What it does |

|---------|-------------|

| `/dc:continue` | Recover your previous session without losing context |

| `/dc:status` | Show current project status |

| `/dc:doctor` | Diagnose and repair framework issues |

| `/dc:capture` | Capture ideas without interrupting your flow |

| `/dc:migrate` | Plan migration between stacks (Vue→React, JS→TS...) |

| `/dc:update` | Update Don Cheli to the latest version |

Reasoning models (15)

| Command | What it does |

|---------|-------------|

| `/razonar:primeros-principios` | Decompose to fundamental truths |

| `/razonar:5-porques` | Iterative root cause analysis |

| `/razonar:pareto` | 80/20 focus |

| `/razonar:inversion` | Solve in reverse: how do I guarantee failure? |

| `/razonar:segundo-orden` | Consequences of consequences |

| `/razonar:pre-mortem` | Anticipate failures before they happen |

| `/razonar:minimizar-arrepentimiento` | Jeff Bezos framework |

| `/razonar:costo-oportunidad` | Evaluate sacrificed alternatives |

| `/razonar:circulo-competencia` | Know the limits of knowledge |

| `/razonar:mapa-territorio` | Model vs reality |

| `/razonar:probabilistico` | Reason in probabilities, not certainties |

| `/razonar:reversibilidad` | Can this decision be undone? |

| `/razonar:rlm-verificacion` | Verification with fresh sub-LLMs |

| `/razonar:rlm-cadena-pensamiento` | Multi-step Context Folding |

| `/razonar:rlm-descomposicion` | Divide and conquer with subagents |

> **📖 Want to see all commands in action with interactive examples?**

> Visit the full guide: **[doncheli.tv/comousar.html](https://doncheli.tv/comousar.html)**

---

## Killer Features

### Drift Detection — Architecture Watchdog

Detects when code diverges from specifications. If code changes but specs don't, you get an immediate alert:

```

⚠️ DRIFT ALERT: Architecture Compromised

  File changed:   src/services/auth.ts (line 45-67)

  Spec affected:  specs/auth/login.feature:23

  Scenario:       "Login with MFA via TOTP"

  Drift:          MFA logic removed, spec still requires it

  Severity:       🔴 CRITICAL

```

```bash

/dc:drift                    # Full project scan

/dc:drift --watch            # Activate async watcher

```

### Time Travel — Reasoning Debugger

See **why** the framework chose each model, skill and decision. Navigate the reasoning history like a timeline:

```

10:15 ─── /dc:start ──────────────────

│  D001: Level detected → 2 (Standard)

│  Model: sonnet (confidence: 85%)

│  Discarded: N1 (>3 files), N3 (1 module)

│

10:22 ─── /dc:specify ────────────────

│  D003: Reasoning → /razonar:pre-mortem

│  Reason: Payment feature (high risk)

│  Model: sonnet → opus (escalated by complexity)

```

```bash

/dc:time-travel              # Full session timeline

/dc:time-travel --adjust     # Tune model thresholds dynamically

```

### Pre-Flight — Cost Simulator

Know **exactly** how much a phase will cost BEFORE executing:

```

┌──────────────────────────────────────────────────┐

│  Fase            │ Tokens   │ Model   │ Cost     │

├──────────────────┼──────────┼─────────┼──────────┤

│  Implement       │ ~45,000  │ sonnet  │ $0.27    │

│  Review          │ ~18,000  │ opus    │ $0.54    │

│  TOTAL           │ ~63,000  │ mixed   │ $0.81    │

└──────────────────┴──────────┴─────────┴──────────┘

✅ Within budget ($0.81 < $5.00)

```

```bash

/dc:preflight                # Estimate pending phases

/dc:preflight --budget 5.00  # Alert if exceeds $5

```

---

## Multi-platform

Don Cheli is not a program. It's Markdown files that any AI agent can interpret.

| Platform | Support | Instruction file |

|----------|---------|------------------|

| **Claude Code** | Full native | `CLAUDE.md` |

| **Google Antigravity** | Native with 5 skills + 4 workflows | `GEMINI.md` |

| **Cursor** | Via universal contract | `AGENTS.md` |

| **Codex** | Via universal contract | `AGENTS.md` |

| **Warp** | Compatible | `CLAUDE.md` |

| **Amp** | Compatible | `prompt.md` |

| **Continue.dev** | Compatible | `AGENTS.md` |

| **OpenCode** | Compatible | `AGENTS.md` |

---

## CI/CD Integration

Enforce quality gates on every Pull Request with one line:

```yaml

# .github/workflows/sdd-check.yml

- uses: doncheli/don-cheli-sdd@main

  with:

    gates: all          # spec, tdd, coverage, owasp, custom

    min-coverage: 85

    comment-pr: true    # posts results as PR comment

```

The action verifies `.dc/` artifacts, TDD compliance, coverage, OWASP and custom gates. [Full CI/CD docs →](docs/ci-cd.md) | [GitLab CI template →](examples/ci/gitlab-ci.yml)

---

## Custom Quality Gates

Define your own rules in `.dc/gates/` as simple YAML files:

```yaml

# .dc/gates/no-console-log.yml

name: No console.log in production

type: grep

pattern: "console\\.log"

files: "src/**/*.ts"

severity: block

```

5 gates included out of the box. Create your own with `/dc:gate create`. [Custom Gates docs →](docs/custom-gates.md)

---

## Telemetry & Dashboard

100% local metrics. No data leaves your machine.

```bash

/dc:metrics             # Summary in terminal

/dc:dashboard           # Interactive HTML dashboard

/dc:dashboard --csv     # Export for corporate reporting

```

Tracks: TDD success rate, coverage trend, quality gates pass rate, estimation accuracy, stubs detected, OWASP findings. [Telemetry docs →](docs/telemetry.md)

---

## SDD Certification

Show that your project was built with engineering discipline. Add these badges to your README:

```markdown

[![SDD Certified](https://img.shields.io/badge/SDD_Certified-Don_Cheli-6c5ce7?style=for-the-badge)](https://github.com/doncheli/don-cheli-sdd)

[![TDD](https://img.shields.io/badge/TDD-Iron_Law_Enforced-00cec9?style=for-the-badge)](https://github.com/doncheli/don-cheli-sdd)

[![OWASP](https://img.shields.io/badge/OWASP-Audited_by_Don_Cheli-e17055?style=for-the-badge)](https://github.com/doncheli/don-cheli-sdd)

```

[![SDD Certified](https://img.shields.io/badge/SDD_Certified-Don_Cheli-6c5ce7?style=for-the-badge)](https://github.com/doncheli/don-cheli-sdd) [![TDD](https://img.shields.io/badge/TDD-Iron_Law_Enforced-00cec9?style=for-the-badge)](https://github.com/doncheli/don-cheli-sdd) [![OWASP](https://img.shields.io/badge/OWASP-Audited_by_Don_Cheli-e17055?style=for-the-badge)](https://github.com/doncheli/don-cheli-sdd)

[Full certification criteria →](docs/certification.md)

---

## Philosophy

> **"Context Window = RAM, File System = Disk"**

1. **Persistence over conversation** — Write it down, don't just say it

2. **Structure over chaos** — Clear files, clear roles

3. **Recovery over restart** — Never lose progress

4. **Evidence over assertions** — Show, don't tell

5. **Simplicity over complexity** — Everything in your language

---

## Community & Support

- [GitHub Discussions](https://github.com/doncheli/don-cheli-sdd/discussions) — Questions and proposals

- [GitHub Issues](https://github.com/doncheli/don-cheli-sdd/issues) — Bugs and feature requests

- [YouTube @doncheli](https://youtube.com/@doncheli) — Tutorials and demos

- [Instagram @doncheli.tv](https://instagram.com/doncheli.tv) — News

- [doncheli.tv](https://doncheli.tv) — Full web documentation

---

## Contributing

See [CONTRIBUTING.md](CONTRIBUTING.md) for the full guide.

---

## License

[Apache 2.0](LICENSE) — Copyright 2026 Jose Luis Oronoz Troconis (@DonCheli)

---



  Stop guessing. Start engineering.



  

  

  



  _{Made with ❤️ in Latin America — Don Cheli SDD Framework}
ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Awesome

https://github.com/doncheli/don-cheli-sdd

Awesome Lists containing this project

README

Don Cheli — SDD Framework
