bybren-llc/safe-agentic-workflow

SAW — SAFe Agentic Workflow AI Agent Harness for Multi-Agent Team Workflows Built on SAFe methodology (Scaled Agile Framework), adapted for AI agent teams (Now With AI-DLC!) Works for any team with repeatable processes: Software, Marketing, Research, Legal, Operations.

What it solves

SAW (SAFe Agentic Workflow) provides a structured harness for coordinating multi-agent AI teams. It prevents the chaos of unstructured AI interactions by implementing a production-tested architecture that enforces quality gates, clear role boundaries, and evidence-based delivery, ensuring that AI agents operate within a professional software development lifecycle (SDLC) rather than as simple chat interfaces.

How it works

The project implements a three-layer architecture to separate concerns:

  1. Hooks: Automatic guardrails that handle format checks and blockers.
  2. Commands: User-invoked workflows (e.g., /start-work, /pre-pr) that automate common tasks.
  3. Skills: Model-invoked domain expertise that the AI loads automatically based on the task.

It adapts the Scaled Agile Framework (SAFe) to AI, mapping specific roles (like Business Systems Analyst or System Architect) to specialized AI agent profiles. It also includes a "Knowledge Vault"—an evidence-verified knowledge base where every claim is linked to a specific commit SHA to detect and prevent knowledge drift.

Who it’s for

Software engineering teams, research, legal, or operations teams that use AI assistants (specifically Claude Code, Gemini CLI, Codex CLI, or Cursor IDE) and require repeatable, structured processes for their work.

Highlights

  • Multi-Provider Support: Compatible with Claude Code, Gemini CLI, Codex CLI, and Cursor IDE.
  • SAFe Integration: Maps professional agile roles to AI agent profiles for structured autonomy.
  • Knowledge Vault: A portable, evidence-verified knowledge base using Open Knowledge Format v0.1 to prevent AI hallucinations and staleness.
  • Dark Factory: Support for persistent autonomous agent teams via tmux on remote servers.
  • Coded Methodology: Implements patterns from six Anthropic engineering research papers.

Related

  • Project
    ZHangZHengEric/Sage

    A production-ready agent platform for task execution and automation that integrates browser workflows and multi-channel IM delivery.

  • Project
    affaan-m/ECC

    An agent harness operating system that provides a coordinated engineering system and toolbox for AI agents to plan, test, and verify code changes.

  • Project
    aden-hive/hive

    A production-grade runtime for colonies of AI agents that manages state, failure recovery, and human oversight to execute reliable business processes at scale.

  • Project
    wshobson/agents

    A marketplace of production-ready agentic building blocks including plugins, agents, and skills that can be deployed across multiple AI coding harnesses like Claude Code and Cursor.

  • Project
    Chorus-AIDLC/Chorus

    A management harness for AI coding agents that implements the AI-DLC workflow, allowing teams to coordinate multiple agents through a structured propose-and-verify pipeline.