me2resh/apexyard

Your AI engineering org, governed. Ship agent-built code safely to production.

ApexYard – AI‑assisted governance for code projects

What it is ApexYard is an operations repository that teams can fork and use to govern a portfolio of software projects that are built (or heavily assisted) by AI coding agents such as Claude Code, opencode, pi, or Codex. It does not run any services; it is just a collection of plain‑text markdown files, Bash hooks, and slash‑command “skills” that the AI agents read and execute.

Why it matters AI can generate working code fast, but moving that code safely into production still requires the usual software‑development safeguards – tickets, reviews, quality gates, migration plans, etc. ApexYard codifies those safeguards in a reusable, tool‑agnostic way, so the same governance rules apply whether you are using Claude Code, opencode, pi, Codex, or even the Cursor IDE.

Key components

  • Roles (≈20) – markdown files that model personas (e.g., Rex the reviewer, Hakim the security guard) and activate on specific triggers.
  • Shell hooks (≈60) – Bash scripts that run on git/gh commands to enforce “ticket‑first” edits, two‑marker merge gates, secret scans, migration checks, etc.
  • Slash‑command skills (≈68) – commands like /setup, /handover, /decide, /code‑review, /migration, /launch‑check that an AI agent can invoke directly from the IDE.
  • Sub‑agents – small agents (Rex, Hakim, Tariq, …) that perform specialised reviews or checks.
  • Configuration files – onboarding.yaml (git‑ignored) holds project‑specific settings; apexyard.projects.yaml tracks the list of repos under governance.

Typical workflow

  1. Fork the ApexYard repo and clone it.
  2. Run /setup (or copy onboarding.example.yaml) to configure your organisation’s policies.
  3. Use /handover <repo> to add a codebase to the portfolio; the command clones the repo, scores its “harnessability”, and records it in apexyard.projects.yaml.
  4. Throughout development, agents invoke skills such as /code‑review <pr> or /decide to record decisions, while the Bash hooks block any git action that violates the ticket‑first or merge‑gate rules.
  5. When a PR is ready, a named human (and optionally the Rex agent) must explicitly approve the exact commit before the merge gate opens.
  6. Periodic /update keeps the fork in sync with upstream ApexYard changes, and a Conformance CI job re‑validates that all supported AI tools still obey the rules.

Tool support

Tool Enforcement Setup step
Claude Code ✅ native (no extra install) just run /setup
opencode ✅ proven (adapter script) bash bin/install-opencode-adapter.sh
pi ✅ proven bash bin/install-pi-adapter.sh
Codex ✅ proven bash bin/sync-codex-adapter.sh
Cursor ✅ native in IDE (loads .claude/) enable third‑party configs + bash bin/install-cursor-adapter.sh

Customization All roles, workflows, and document templates are plain markdown files, so teams can add new roles, edit existing workflows, or drop in additional templates without touching any compiled code.

Getting started The README walks you through the six‑step quick‑start (fork → clone → add upstream → /setup → /handover → start using portfolio commands). A status line plugin (apexyard-status) can be added to the Claude Code prompt to show the active ticket and pending approvals.

Who might use it

  • Companies that let AI agents write code but need a disciplined SDLC.
  • Teams that manage many micro‑services or front‑end apps and want a single source of truth for policies.
  • Developers who prefer a “no‑service”, file‑based governance layer that works across multiple AI coding assistants.

All details above are taken directly from the repository’s README; no additional features have been inferred.

Related

  • Project
    ai-driven-dev/framework

    AI‑Driven Dev Framework (AIDD) – open‑source plugins that embed a full software‑development life‑cycle into AI coding assistants (Claude Code, Cursor, Copilot, etc.). Install six stable plugins, run an onboarding command, then issue high‑level slash commands to plan, implement, validate, review and ship features automatically.

  • Project
    Chachamaru127/claude-code-harness

    A management framework for AI coding assistants that adds structured planning, safety guardrails, and mandatory review steps to ensure reliable code generation.

  • Project
    ai-sdlc-framework/ai-sdlc

    AI-SDLC is an orchestration framework that turns well-specified contracts into deterministic AI-driven software delivery. It uses a Definition-of-Ready gate, a session hierarchy of subagents, cross-harness review, and attestation to govern how AI agents work on mature codebases.

  • Project
    iamfakeguru/agent-md

    agent‑md is a lightweight framework that installs rule files, git‑hooks, and helper scripts to turn AI coding assistants (Claude Code, Codex, Cursor, Windsurf) into contract‑driven agents. It provides a single source‑of‑truth markdown file (AGENT.md) plus enforceable checks (type‑check, lint, tests, visual screenshots, state updates) that block the model until the required evidence is present.

  • Project
    HoangNguyen0403/agent-skills-standard

    A portable SDLC standards layer for AI coding agents that turns engineering rules into modular, version-controlled skills to reduce token usage and ensure consistent code quality.