# Architecture

Current deployed architecture. Updated in the same PR as any structural change.

## Layered model

```
this repo (global defaults)
    ├── install.sh → ~/.agents/skills/  (canonical skills location)
    ├── install.sh → ~/.claude/skills/  (symlink → ~/.agents/skills/, Claude Code adapter)
    └── install.sh → ~/.claude/         (Claude Code config + content)

project repo (local overrides)
    └── .claude/settings.json, CLAUDE.md (overrides global)
```

Projects consume from this repo by pulling updates via `sync.sh` (Chunk 6). Until then, install is a one-time manual step.

## Directory structure

```
ai-development/
├── AGENTS.md                    # Provider-agnostic always-on rules for this repo; imported by repo CLAUDE.md
├── CONTEXT.md                   # Domain language, principles, glossary; auto-loaded at session start
├── docs/                        # Workflow artifacts and issues (prd/, ard/, bug/, notes/, adr/, issues/, spec/) + research/ (raw research audit trail)
├── .agents/                     # Agent Skills standard location (provider-agnostic)
│   ├── skills/                  # SKILL.md files — canonical source, deployed to ~/.agents/skills/
│   └── evals/                   # eval.yaml files for skills not yet in a plugin (cross-cutting/, implement/)
├── .claude-plugin/              # Marketplace manifest (both Claude Code and Copilot CLI read here)
│   └── marketplace.json         # Declares all installable plugins in this repo
├── .github/plugin/
│   └── marketplace.json         # Mirror of .claude-plugin/marketplace.json for Copilot CLI canonical path
├── plugins/                     # Installable plugin units (each is a self-contained deployable)
│   └── kyberforge/              # Marketplace management toolkit (create-plugin, marketplace-architect, write-skill, write-eval)
├── core/                        # Provider-agnostic source of truth
│   ├── AGENTS.md                # Global always-on rules (Communication + Behavior); deployed to ~/.agents/AGENTS.md
│   ├── instructions/            # AI behavior definitions (plain markdown)
│   ├── agents/                  # Agent role definitions
│   ├── workflows/               # Workflow definitions
│   └── prompts/                 # Reusable prompt templates
├── providers/                   # Provider-specific adapters
│   ├── claude-code/             # CLAUDE.md (thin adapter), settings.json, provider-manifest.sh
│   └── copilot/                 # copilot-instructions.md, hooks, agents adapter
└── scripts/
    ├── deploy-manifest.sh       # Source→target mappings; sourced by install.sh and sync.sh
    ├── install.sh               # Deploys to ~/.agents/skills/, ~/.claude/, etc.
    ├── sync.sh                  # Pulls updates into an existing project
    └── init-project.sh          # Bootstraps a new or existing project
```

## Content deployment model

`install.sh` is a **deployer**, not a composer. It does not concatenate content into a single file. Instead:

- `.agents/skills/` → `~/.agents/skills/` — canonical skills location; each skill dir is replaced individually (parent not wiped, user-added skills preserved)
- `core/AGENTS.md` → `~/.agents/AGENTS.md` — global always-on rules (Communication + Behavior); imported by `~/.claude/CLAUDE.md` via `@~/.agents/AGENTS.md`
- Provider adapters declared in `providers/*/provider-manifest.sh` — symlinks from the provider's skill path to `~/.agents/skills/`; e.g. Claude Code gets `~/.claude/skills/ → ~/.agents/skills/` because it reads `~/.claude/skills/` natively. Providers that read `~/.agents/skills/` directly need no adapter.
- `core/` → `~/.claude/core/` — workflows, prompts, agent definitions; agent reads on demand
- `providers/claude-code/settings.json` → `~/.claude/settings.json`
- `providers/claude-code/CLAUDE.md` → `~/.claude/CLAUDE.md` — thin adapter: imports `~/.agents/AGENTS.md` and `governance.md`; no original content

`~/.claude/CLAUDE.md` is a thin adapter, not a content source. It imports `~/.agents/AGENTS.md` (always-on rules) and `governance.md` (always-on governance), then lists the content index. All always-on content lives in `AGENTS.md` files so other providers can import the same source without duplication.

## Governance layer

`core/instructions/governance.md` is the always-on governance instruction file. Unlike the on-demand instruction files in the content index, governance.md is loaded into every Claude session via `@import` in `providers/claude-code/CLAUDE.md`. This is a technical guarantee, not a behavioural instruction — `@import` causes Claude Code to expand and load the file at launch, before any interaction begins.

The governance layer has two phases:
- **Phase 1** (complete): instruction and documentation layer — `governance.md` loaded via `@import`; `docs/ai-constitution.md` and `docs/HUMANS.md` as human-facing reference; `CONTEXT.md` extended with governance domain language.
- **Phase 2** (Chunk 6): deterministic enforcement layer — pre-commit hooks, CI gates, secret scanning, licence scanning. Specified in `docs/research/governance_principles/CONTROLS.md`.

## AGENTS.md pattern

This repo uses two `AGENTS.md` files as the provider-agnostic source of always-on rules (ADR-0012):

- **Repo-level `AGENTS.md`** — instructions for agents working inside this repo (structure, key rules, chunk workflow). Imported by repo `CLAUDE.md` via `@AGENTS.md`.
- **Global `core/AGENTS.md`** — Communication and Behavior rules that apply across all projects. Deployed to `~/.agents/AGENTS.md`; imported by `~/.claude/CLAUDE.md` via `@~/.agents/AGENTS.md`.

Both `CLAUDE.md` files are thin adapters: they import from their respective `AGENTS.md` and add only Claude Code-specific syntax (`@import`, content index paths). They carry no original always-on content.

This repo also has a `CLAUDE.md` at its root — the Claude Code entry point for working in this repo. It imports `AGENTS.md` and `CONTEXT.md`, nothing more. This is distinct from `providers/claude-code/CLAUDE.md`, which is the global config deployed to `~/.claude/`.

## Provider model

`core/` is never tool-specific. `providers/` is never shared. When adding a new provider, write an adapter in `providers/<name>/` that translates core content into the tool's expected format and location. The core content itself does not change.

Skills are the strongest shared primitive — the `SKILL.md` format and [Agent Skills open standard](https://agentskills.io) are cross-provider. Providers that don't read `~/.agents/skills/` natively declare a symlink adapter in `providers/<name>/provider-manifest.sh`; `install.sh` discovers and wires these up automatically.

## Architectural decisions

Key hard-to-reverse decisions are recorded as ADRs in `docs/adr/`. See the index there for rationale on choices like the pull distribution model, copy-not-symlink coupling, and the two-tier CLAUDE.md structure.