Skip to content

Releases: vagnerfirminopro/v-cos

V-COS v1.0 — Governance layer for AI-native projects

14 Jun 20:53
@vagnerfirminopro vagnerfirminopro
v1.0.0
f485478

Choose a tag to compare

View all tags
V-COS v1.0 — Governance layer for AI-native projects Latest
Latest

V-COS v1.0

The first public release of V-COS — the governance layer that keeps AI-native projects coherent over time. Extracted from a production SaaS built by a solo founder in ~2 months, for under $100 in costs, across 25+ development cycles with AI as co-developer.

V-COS wasn't designed in the abstract. It was born out of necessity — a solo founder building a real product with limited resources, distilled from solving, one by one, the governance, orchestration, architecture, and context problems that show up when you operate an AI-native project with no team and no budget.

What's in this release

The five problems V-COS solves
Context decay between sessions · implicit hierarchy · cognitive content mixed with technical reference · no self-evaluation · no session protocol.

The three layers

  • Layer 1 — Document Governance: explicit hierarchy, session protocol, hot memory, self-evaluation.
  • Layer 2 — Skills Architecture: IDE skills, squad skills, taxonomy.
  • Layer 3 — Agent Governance: design, specialization, orchestration.

Layers 1 and 2 are model-agnostic. Layer 3 ships with a reference implementation for Claude Code.

What's in the repo

  • docs/ — the framework in depth (five problems, three layers, harness engineering, spec-driven development)
  • templates/ — copy-and-adapt files (AGENTS, CLAUDE, CYCLE-CURRENT/ARCHIVE, REFERENCE, decisions, reference sub-agents)
  • skills/ — a starter skill pack, agnostic and ready to adapt
  • implementations/ — how to wire it to Claude Code, Cursor, or any tool
  • examples/v-strategy/ — the real (sanitized) project V-COS was extracted from
  • PRINCIPLES.md — the eight governance principles

Getting started

The minimum viable setup is four files and runs in ~1h. Copy the templates from templates/, adapt them to your project, and point your agent at AGENTS.md as the first file it reads.

Why open source

Out of purpose — to help more people create, build, and maintain projects with AI over time — and out of pragmatism: V-COS is a living system, and with the community it evolves faster, with more quality per contribution and more use cases. Contributions, issues, and PRs are welcome.

License

MIT.

Full Changelog: https://github.com/vagnerfirminopro/v-cos/commits/v1.0.0

Assets 2
Loading