coleam00/ai-software-factory

What it solves

AI Software Factory creates an autonomous software development lifecycle (SDLC) that transforms GitHub issues or a Product Requirements Document (PRD) into validated, merged code without human intervention. It focuses on the "dark factory" concept—lights-out manufacturing for software—where the primary challenge is establishing enough trust in automated merges to remove the human from the keyboard.

How it works

The system leverages a pinned version of Archon's SDLC pack to execute a series of shared workflows. It operates by:

1. Intake: Converting a PRD into a backlog of issues via archon-backlog or picking up existing GitHub issues.
2. Implementation: Using archon-ship or archon-lifecycle to plan, implement, and review code changes.
3. Verification: Running end-to-end journeys (defined in END-TO-END.md) and independent holdout scenarios (HOLDOUT.md) to ensure the code behaves as expected.
4. Merging: Using a merge queue to merge validated code into the default branch based on configured approval modes.
5. Deployment: Optionally rolling the default branch onto a running service via archon-deploy.

Who it’s for

Developers and product owners who want to automate the entire path from requirement to deployment, specifically those who can define a clear product mission and validation journeys to guide an AI agent.

Highlights

• Autonomous Backlog Generation: Slices a PRD into engineering-ready issues automatically.
• Evidence-Based Validation: Uses runtime verification and holdout tests to prevent regressions and ensure correctness.
• Mission-Driven Guardrails: Uses a MISSION.md file to define what a product must never become, preventing AI-driven feature drift.
• Server-Ready: Includes tooling for deployment as a systemd service for persistent, unattended operation.
• Integrated SDLC: Orchestrates triage, planning, implementation, review, and deployment through a shared workflow pack.