Civyk Hatch

"Build fair. Build clean. Build for people."

Zero-prompt architecture CLI for AI-powered development workflows — Progressive prompt injection that guides AI agents through structured development stages.

Problem: AI coding assistants are powerful but unpredictable. Without structure, they produce inconsistent results, skip edge cases, and lack traceability from requirements to code.

Solution: Civyk Hatch provides structured, stage-based workflows that guide AI agents through specification, planning, implementation, and review — with full state persistence and auto-resume.

---

Core Features

• Progressive Prompt Injection — Small, focused prompts (50-80 lines) at each stage
• Folder-Based State Management — Simple state persistence using feature folders
• Template System — Reusable templates with variable substitution
• Multi-Workflow Support — Specification, planning, implementation, review, and more
• Cross-Platform — Works on Linux, macOS, and Windows
• Free binaries — Compiled binaries available via PyPI at no cost

---

Quick Start

Installation

pip install civyk-hatch

Setup for Your AI Agent (Optional)

Slash commands are optional. You can use direct prompting without any setup (see Usage).

To install for your AI agent:

cd /path/to/your/project

# Fully interactive setup (recommended)
civyk-hatch init

# CLI mode: slash commands
civyk-hatch init . --ai claude --mode cli

# MCP mode: JSON-RPC server
civyk-hatch init . --ai claude --mode mcp

# Configure all supported agents at once
civyk-hatch init . --all

The interactive mode prompts for project location, AI assistant, and integration mode.

Supported agents: claude, cursor-agent, windsurf, copilot, opencode, kilocode, antigravity, gemini, qwen, codex, roo, amp, q

Verify Installation

civyk-hatch --version

---

Available Workflows

Command	Description	Stages
constitution	Create project constitution	3
specify	Create baseline specification	6
clarify	Ask structured questions	3
plan	Create implementation plan	4
tasks	Generate actionable tasks	4
analyze	Cross-artifact consistency analysis	3
implement	Execute implementation	5
tests	Iterative test implementation	2
review	Deep code review and remediation	2
checklist	Generate quality checklist	3
analyze-project	Analyze existing project	16
deepwiki	Generate AI-powered wiki documentation	16
generate-guidelines	Generate coding guidelines from documents	4
sync-docs	Sync documentation with codebase	2
understand	Build hierarchical codebase understanding	5
orchestrate	Orchestrate complete spec-driven workflow	-

---

Usage

List Available Commands

civyk-hatch --help

Option 1: Direct Prompting (No Setup Required)

Simply prompt your AI coding assistant directly:

Run CLI: `civyk-hatch specify`. Follow all instructions in the output.

Replace specify with any available command:

Run CLI: `civyk-hatch review`. Follow all instructions in the output.
Run CLI: `civyk-hatch tests`. Follow all instructions in the output.
Run CLI: `civyk-hatch plan`. Follow all instructions in the output.
Run CLI: `civyk-hatch analyze-project`. Follow all instructions in the output.

The AI agent will execute the CLI command and follow the emitted prompts automatically.

Option 2: Slash Commands (Requires Init)

For a more integrated experience, run civyk-hatch init to install slash commands:

/civyk-hatch.specify     # Start specification workflow
/civyk-hatch.plan        # Start planning workflow
/civyk-hatch.implement   # Start implementation workflow
/civyk-hatch.review      # Start code review workflow
/civyk-hatch.tests       # Start test implementation workflow

How slash commands work:

1. User types slash command in their AI assistant
2. Slash command loads a prompt that instructs the AI agent
3. AI agent executes civyk-hatch CLI commands to drive the workflow
4. CLI emits stage prompts that guide the AI through each step

Auto-Resume

All feature-scoped commands support automatic state detection:

civyk-hatch plan --stage=1     # Early stages need explicit --stage
civyk-hatch plan --stage=2     # Still no state (created at stage 3)
civyk-hatch plan               # Stage 3+: auto-detects everything!
civyk-hatch resume             # See what's next

---

Integration with civyk-repoix

Civyk Hatch integrates with civyk-repoix for enhanced code intelligence:

• MCP Mode — Direct integration when MCP is available
• CLI Mode — Fallback for restricted environments

The review and test workflows automatically detect which mode is available and use appropriate commands.

---

Configuration

AGENTS.md

Place AGENTS.md in your repository root to configure AI agent behavior:

my-project/
├── AGENTS.md          # AI agent guidelines (repo root only)
├── memory/
│   └── constitution.md
└── specs/

The AI agent reads AGENTS.md at workflow start for project-specific instructions.

---

Principles

This project follows Civyk's core values:

1. Users are customers, not inventory — We make money by providing value
2. No dark patterns, ever — Honest communication, no manipulation
3. No silent telemetry — You know exactly what we collect and why
4. Consent is explicit & reversible — Full control over your data
5. Long-term trust > short-term growth — Building for sustainability

---

Support

Help keep this project alive and growing!

If Civyk Hatch has helped your development workflow, consider supporting its continued development. Your contribution helps with:

• Ongoing maintenance and bug fixes
• New feature development
• Infrastructure costs

50% of all donations go directly to children's charities helping those in need. The remaining funds support project maintenance and feature upgrades.

Every contribution, no matter the size, makes a difference.

---

License

Proprietary — see LICENSE

Free to use: Compiled binaries are available via PyPI at no cost for personal and commercial use.

---

Built on Trust, Driven by Value