About this project

AI-SDLC is a decision engine for spec-driven AI workflows, covering the contract-to-shipped half of a spec-driven development stack. It addresses the problem that AI agents can build small projects but software quality degrades as codebases grow, by frontloading operator decisions through a Definition-of-Ready gate and executing well-specified contracts deterministically. The framework consists of five pillars: a Decision Engine with a Definition-of-Ready gate and forthcoming Decision Catalog; an Autonomous Pipeline Orchestrator that walks dependency graphs, dispatches tasks into isolated git worktrees, and runs a Step 0-13 pipeline; Cross-Harness Review with three parallel reviewer subagents and DSSE attestations enforcing independence by construction; an Operator TUI with five panes for monitoring decisions, pipelines, dependency graphs, configuration, and analytics; and Declarative Governance with resources like Pipeline, Decision, AgentRole, QualityGate, AutonomyPolicy, and AdapterBinding, all backed by JSON Schema. The project is agent-agnostic, supporting Claude Code, Codex, Cursor, Copilot, Aider, or any OpenAI-compatible API. It ships as multiple packages including an orchestrator runtime, pipeline CLI, Claude Code plugin, TypeScript/Python/Go SDKs, and a conformance test suite. The specification follows Kubernetes-style API maturity, currently at v1alpha1. The project is licensed under Apache 2.0.