Coding & Developmentโœจgemini๐Ÿค–claude๐ŸŒuniversal

Build backend

GitCortex delivers production-grade software directly from social media conversations. It actively resolves ambiguity, architects solutions, and orchestrates multiple AI coding agents (Claude Code, Gemini CLI, Codex, etc.) in parallel with built-in three-layer quality gates. No IDE, no terminal, no manual intervention.

Quick Start

Install this Skill

npx skills add huanchong-99/SoloDawn
<p align="center">
  <img src="installer/assets/solodawn.ico" alt="SoloDawn" width="120" />
</p>

<h1 align="center">SoloDawn</h1>

<p align="center">
  <strong>Give it one sentence. It builds the entire project. You just wait for delivery.</strong>
</p>

<p align="center">
  <a href="README.zh-CN.md">็ฎ€ไฝ“ไธญๆ–‡</a>
</p>

---

## What Is SoloDawn?

SoloDawn's ultimate design goal is to **complete complex, production-grade products through a simple conversation on a social platform** โ€” not toy demos, but real, complex, production-ready software.

> **One-liner:** Whether you're a programmer or not, just describe what you need. Everything else is automatic.

---

## The Problem We Solve

There's a fundamental contradiction in AI-assisted coding today:

**For programmers** โ€” you use vibe coding, but you still have to set up workflows, configure skills, wire up MCP servers, write plans, organize docsโ€ฆ a mountain of prerequisite work before you get satisfying results. At its core, the human is still the driver.

**For non-programmers** โ€” say, a project manager โ€” they can't even translate a requirement into a technical spec. The reality is: AI can only boost your productivity if you already know how to do the work yourself.

**SoloDawn solves both problems:**

| Who you are | What you do | What happens |
|---|---|---|
| **Programmer** | Throw in the task goal | Fully automatic development. No commands to issue, no "Continue" to click. Just wait for delivery. Your existing workflows, skills, MCP servers? All compatible. Zero migration cost. |
| **Non-programmer** | Describe what you want in plain language | The system proactively asks follow-up questions in plain language until everything is clear, then automatically generates a formal technical spec on the backend. After that โ€” same deal โ€” just wait for delivery. |

**The best partner for a one-person company.** With this project, you essentially have an entire professional development team.

---

## Two Work Modes

### 1. Manual Workflow (Advanced)

For power users who want full control:

- Choose how many terminals to use and their roles
- Select the model for each terminal
- Configure which skills, MCP servers, and slash commands each terminal uses
- Full customization over every aspect of the workflow graph

### 2. Orchestrated Workspace

The orchestrated workspace has two sub-modes:

#### Direct Execution Mode
For technical users โ€” provide a detailed, specific requirement and the system executes it directly. No questions asked.

#### Guided Conversation Mode
For non-technical users โ€” provide a vague idea and the system will:
1. **Proactively ask follow-up questions in plain language** until the requirement is fully understood
2. **Automatically generate a formal technical specification** on the backend
3. **The upper-level orchestrator Agent takes over**, autonomously commanding all terminals until the product is delivered

---

## Built-in Quality Gates โ€” Solving AI Hallucination at Scale

AI-generated code has a quality problem. SoloDawn addresses this head-on with a **three-layer quality gate system** that maximally mitigates AI hallucination in code output:

| Gate | Trigger | Scope |
|------|---------|-------|
| **Terminal Gate** | Every checkpoint commit | Changed files only โ€” cargo check, clippy, tsc, tests, secret detection |
| **Branch Gate** | Last terminal in a task completes | Full task branch โ€” all checks + lint + coverage + complexity |
| **Repo Gate** | Before merge to main / CI | Full repo โ€” all checks + SonarQube analysis + security scan |

**Four enforcement modes:** `off` โ†’ `shadow` โ†’ `warn` โ†’ `enforce`

When a quality gate fails, structured fix instructions are automatically sent back to the same terminal. The terminal fixes the issues and re-commits โ€” no human intervention required. This creates a **self-healing development loop** that catches and corrects hallucination-induced errors before they reach your codebase.

---

## Core Design Philosophy

- **Upper-layer orchestration, not code generation.** The orchestrator Agent never writes code โ€” it commands the best professional AI CLIs (Claude Code, Gemini CLI, Codex, Amp, Cursor Agent, etc.) to do the work.
- **Non-invasive by design.** SoloDawn doesn't replace any CLI, modify any config, or define new tools. It inherits the full native ecosystem of every CLI โ€” all slash commands, plugins, skills, and MCP servers work unchanged. Your existing setup? Zero migration cost.
- **Git-driven event loop.** The orchestrator only consumes LLM tokens when a Git commit event occurs. Between events, it sleeps at zero cost โ€” saving 98%+ tokens compared to polling.

---

## Architecture

```
           โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”
           โ”‚        Orchestrator Agent (LLM-driven)       โ”‚
           โ”‚      Dispatches ยท Monitors ยท Merges          โ”‚
           โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜
                                 โ”‚
            โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ผโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€

... (truncated)
Repository
165stars
12forks
View on GitHub
Details
DiscoveredApr 15, 2026
Last UpdatedApr 15, 2026
Quality Score
60/100