# Documentation

- [npm publishing](NPM-PUBLISHING.md) — package boundary, trusted publisher, staged approval, and recovery

AI Foundry Desk documentation is organized by what you are trying to accomplish. Start with a task
guide; use the concept and reference material when you need the underlying model or exact contract.

## Start here

| Audience or task | Recommended path |
| --- | --- |
| New user | [Project overview and quick start](../README.md) → [Layer 1](README-LAYER-1.md) → [Layer 2](README-LAYER-2-AGENTS.md) |
| Returning user | [Terminal UI](TUI.md) or [CLI reference](CLI.md) → [environment ownership](ENVIRONMENT-OWNERSHIP.md) |
| Skill/profile maintainer | [Skills and profiles](LAYER-2-SYNC.md) → [Layer 3 recipes](LAYER-3-RECIPES.md) |
| MCP configuration maintainer | [MCP configuration](MCP-CONFIGURATION.md) → [CLI reference](CLI.md#manage-mcp-configuration) |
| Repository maintainer | [Project harnesses](PROJECT-HARNESSES.md) → [validation matrix](VALIDATION-MATRIX.md) |
| Operator | [Observability](OBSERVABILITY.md) → [sandbox repair](AGENT-SANDBOX-REPAIR.md) |
| Contributor | [Contributing](../CONTRIBUTING.md) → [development](DEVELOPMENT.md) → [architecture](ARCHITECTURE.md) |

## Task guides

- [Use the interactive terminal interface](TUI.md)
- [Install and verify Layer 1](README-LAYER-1.md)
- [Install and verify Layer 2](README-LAYER-2-AGENTS.md)
- [Synchronize skills and profiles](LAYER-2-SYNC.md)
- [Manage scoped MCP configuration](MCP-CONFIGURATION.md)
- [Plan and apply Layer 3 recipes](LAYER-3-RECIPES.md)
- [Audit and govern project instructions](PROJECT-HARNESSES.md)
- [Operate local observability](OBSERVABILITY.md)
- [Repair sandbox tool access](AGENT-SANDBOX-REPAIR.md)
- [Maintain backups](ENVIRONMENT-OWNERSHIP.md#backup-ownership)

## Understand the system

- [Architecture](ARCHITECTURE.md): modules, boundaries, state, and platform adapters
- [Layer inventory](LAYER-INVENTORY.md): installed products, purposes, owners, and platforms
- [Platform support](PLATFORM-SUPPORT.md): validated environments and support limits
- [Environment ownership](ENVIRONMENT-OWNERSHIP.md): what AFD manages and what it leaves alone
- [Security boundaries](SECURITY-BOUNDARIES.md): trust model and protected data
- [Observability design](OBSERVABILITY-PLAN.md): implementation rationale and evidence model
- [MCP configuration design](MCP-CONFIGURATION-DESIGN.md): registry contracts, adapter architecture, and rationale

## Reference and project information

- [Terminal UI](TUI.md): interactive taxonomy, workflow, keymap, safety, accessibility, and real screens
- [CLI reference](CLI.md): every public and operational `afd` command
- [Validation matrix](VALIDATION-MATRIX.md): agent and environment coverage
- [Roadmap](ROADMAP.md): current direction and deliberately deferred capabilities
- [Development](DEVELOPMENT.md): local setup and checks
- [Release process](RELEASING.md): build, audit, and publication procedure
- [Changelog](../CHANGELOG.md): released changes

`NEXT-SESSION.md` and `TYPESCRIPT-PLATFORM-MIGRATION.md` are maintainer history and handoff records,
not starting points for users.