# AI Development Auto-Pipeline

A 13-phase automated pipeline that turns a task description into reviewed, tested and committed code, using Claude Code or Codex with adversarial design and security review gates between each phase.

AI Development Auto-Pipeline runs a task description through 13 phases covering design review, security scanning, testing and a final code-review-and-commit gate, so a coding agent's output is checked before it ships rather than after. Each phase prompt follows a fixed constraints, context, task, format, verify structure.

It reviews the design adversarially, scans for OWASP issues, runs deterministic checks before several phases so a clean run makes no model calls, and reviews the actual git diff before committing, auto-healing up to twice on a rejected review. It works with Codex and Claude Code through one Bash engine, and it is MIT licensed.

Website: https://github.com/TheAstrelo/Claude-Pipeline