Skip to content

TAXOVELA (ArchHarness)

Repository: axisvela/TAXOVELA · License: MIT · Language: Python · Layer: Design and standards

An AI-native architecture practice: requirements, design, diagrams, validation, and a deterministic gate.

The problem

Enterprise architecture work is often a pile of documents and diagrams that disagree with each other and cannot be checked. AI assistants can generate architecture artifacts quickly, but speed without a gate produces confident, unverifiable output.

TAXOVELA treats architecture as a pipeline with artifacts, hashes, and gates — not as a chat transcript.

What it provides

  • A set of skills for Claude Code, OpenCode, Codex, GitHub Copilot, and Cursor: requirements, design, diagram generation, validation, security review, optimization, and reporting.
  • A real CLI (python -m archharness) with a multi-project workspace.
  • A workflow gatekeeper that enforces stage order and required artifacts.
  • Hash-verified artifacts: a diagram cannot be edited underneath a recorded decision.
  • Six-dimension diagram validation and a CI enforcement gate with PASS/WARN/BLOCK outcomes.
  • Two renderers from one model: editable draw.io source and presentation-ready D2 output.

Why it matters

  • Deterministic evidence. The gate is reproducible; the output is not just “the model said it looked fine.”
  • Works with the tools people already use. The same architecture practice runs across five AI coding assistants.
  • Designed for review. Requirements, design, validation, security, and committee review are separate, inspectable stages.

How it fits

TAXOVELA is design-time governance. It complements the runtime and control-plane projects by making the architecture around them reviewable. It is adjacent to PAMP (portfolio management) and is used with Enterprise Autonomy reference architectures.

Related but separate: ea-harness is an earlier public architecture skill-pack repository. TAXOVELA is the current AI-native project.

Start