@agents-inc/cli 0.47.0 → 0.50.0

package/CHANGELOG.md

@@ -9,6 +9,45 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ---
 
+## [0.50.0] - 2026-02-27
+
+**Pre-1.0 cleanup — agent.yaml rename, expert mode removal, scroll hook extraction, DRY consolidation**
+
+- Rename `agent.yaml` to `metadata.yaml` across all 17 agents
+- Remove `cli-migrator` agent and expert mode toggle
+- Extract `useSectionScroll` and `useRowScroll` hooks from 6 wizard components
+- Unify scroll rendering, fix domain sort order stability (B-07)
+- Dashboard view when `init` is run on existing project
+
+See [changelogs/0.50.0.md](./changelogs/0.50.0.md) for full details.
+
+---
+
+## [0.49.0] - 2026-02-27
+
+**Config file generation for new skill and new marketplace commands**
+
+- `new marketplace` scaffolds `config/skill-categories.yaml` and `config/skill-rules.yaml`
+- `new skill` creates or updates config files when running in marketplace context
+- Matrix decomposition (D-50) all 8 phases complete
+
+See [changelogs/0.49.0.md](./changelogs/0.49.0.md) for full details.
+
+---
+
+## [0.48.0] - 2026-02-27
+
+**Matrix decomposition, displayName rename, E2E test infrastructure**
+
+- Decompose `skills-matrix.yaml` into `skill-categories.yaml` + `skill-rules.yaml` with directory-based skill discovery
+- Rename `cliName` to `displayName` in metadata schema, loaders, and validators
+- Remove per-skill `categoryExclusive` — exclusivity now driven by category definitions
+- Full PTY-based E2E test suite with 264 tests covering all commands and interactive flows
+
+See [changelogs/0.48.0.md](./changelogs/0.48.0.md) for full details.
+
+---
+
 ## [0.47.0] - 2026-02-25
 
 **Custom domain support, source validation, wizard improvements, dead code cleanup**

package/README.md

@@ -4,190 +4,185 @@
 
 # Agents Inc
 
-An agent composition framework that builds stacks and compiles specialized subagents for [Claude Code](https://docs.anthropic.com/en/docs/claude-code).
+An agent composition framework that builds stacks and compiles specialized subagents for [Claude Code](https://docs.anthropic.com/en/docs/claude-code). Manage your subagents and skills with code via a unified CLI.
 
 [![npm version](https://img.shields.io/npm/v/@agents-inc/cli)](https://www.npmjs.com/package/@agents-inc/cli)
 [![TypeScript](https://img.shields.io/badge/TypeScript-strict-blue.svg)](https://www.typescriptlang.org/)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](./LICENSE)
 [![Node: 18+](https://img.shields.io/badge/Node-18%2B-green.svg)](https://nodejs.org/)
 
-```bash
-npx @agents-inc/cli init
-```
-
-<!-- TODO: Add animated GIF of the full init wizard flow (stack selection → skill grid → agent config → done).
+<!-- TODO: Add animated GIF of the full init wizard flow (stack selection -> skill grid -> agent config -> done).
 This is the hero image. It should show the complete happy path in ~15-20 seconds.
 <p align="center">
   <img src="./assets/init-wizard.gif" alt="Agents Inc init wizard" width="700">
 </p>
 -->
 
-## What this does
+## Overview
 
-Claude Code doesn't know your stack. It doesn't know your patterns, your conventions, or the specific way you use your tools. So you end up repeating the same instructions or maintaining freeform markdown skills that fail silently when something's wrong.
+Agents Inc comprises two parts:
 
-Agents Inc fixes this with structured skills: focused knowledge modules for specific technologies. Each skill covers patterns, conventions, anti-patterns, edge cases, and real code examples. Skills get compiled into specialized subagents (a web developer, a reviewer, a tester, a PM) that actually know what they're doing.
+- **[Marketplace](https://github.com/agents-inc/skills):** 87+ atomic, best-practice skills spanning web, API, CLI, infra, and meta domains
+- **CLI:** installs, ejects, creates, compiles, and updates skills and subagents
 
-The whole thing is opt-in and works alongside your existing Claude Code setup. It uses smart defaults, but there are progressive levels of customization: edit the config, swap skills, choose which to preload and which to load dynamically, eject templates, eject agent partials, or eject entire skills.
+Skills are composed into 18+ domain-specific subagents through templates and modular agent partials, all validated against strict Zod schemas. Everything is ejectable so there's endless customisation with no lock-in.
 
-Under the hood, the CLI is written in strict TypeScript with Zod for runtime validation and JSON Schema generation. Skills and agents are validated against these schemas, so misconfigurations surface immediately. Agent compilation uses Liquid templates, so the output format is fully customizable without touching CLI internals.
+## How It Works
 
-## Getting started
+Each subagent is composed from modular partials (role, workflow, output format) plus a set of atomic skills. The CLI manages the full lifecycle:
+
+- **Install:** pull skills from the public marketplace, your own, or a local directory either as plugins or directly as source files
+- **Compose:** map skills to subagents, configure preloaded vs dynamic loading
+- **Compile:** resolve skill mappings, Liquid templates, and agent partials into `.claude/agents/`
+- **Eject:** take progressive ownership of any layer (partials, templates, skills, subagents)
+- **Update:** pull upstream skill changes without losing local customizations
+
+## Getting Started
 
 ```bash
-# Run the wizard
 npx @agents-inc/cli init
-
-# Or install globally
-npm install -g @agents-inc/cli
 ```
 
 Requires Node 18+ and [Claude Code](https://docs.anthropic.com/en/docs/claude-code).
 
-The wizard walks you through the full setup:
-
-### 1. Pick a stack or start from scratch
+The wizard has three steps:
 
-<!-- TODO: Screenshot of the stack selection screen.
-Shows the list of available stacks (nextjs-fullstack, angular-stack, vue-stack, etc.)
-with descriptions and skill counts.
-<img src="./assets/screenshot-stacks.png" alt="Stack selection" width="700">
--->
-
-Stacks bundle skills together with pre-configured agents. Instead of picking skills individually, grab a stack that matches your setup:
-
-- **nextjs-fullstack**: Next.js + React + Hono + Drizzle + PostHog + Zustand + React Query
-- **angular-stack**: Angular 19 + Signals + NgRx SignalStore + Hono + Drizzle
-- **vue-stack**: Vue 3 Composition API + Pinia + Hono + Drizzle
-- **nuxt-stack**: Nuxt + Vue 3 full-stack + Pinia + Hono + Drizzle
-- **remix-stack**: Remix + React + Hono + Drizzle
-- **solidjs-stack**: SolidJS + Hono + Drizzle
-- **react-native-stack**: React Native + Expo + Zustand + React Query
-- **meta-stack**: Agents for creating agents, skills, docs, and extracting patterns
-
-Or select "Start from scratch" to browse domains and pick skills one by one.
+**1. Pick a stack or start from scratch** (or create your own in your custom marketplace):
 
-### 2. Customize skills
+| Stack              | Technologies                                                       |
+| ------------------ | ------------------------------------------------------------------ |
+| `nextjs-fullstack` | Next.js + React + Hono + Drizzle + PostHog + Zustand + React Query |
+| `angular-stack`    | Angular 19 + Signals + NgRx SignalStore + Hono + Drizzle           |
+| `vue-stack`        | Vue 3 Composition API + Pinia + Hono + Drizzle                     |
+| `nuxt-stack`       | Nuxt + Vue 3 full-stack + Pinia + Hono + Drizzle                   |
+| `remix-stack`      | Remix + React + Hono + Drizzle                                     |
+| `solidjs-stack`    | SolidJS + Hono + Drizzle                                           |
 
-<!-- TODO: Screenshot of the skill selection grid.
-Shows skills organized by category (web, api, mobile, etc.) with checkboxes,
-tagged by source when multiple sources are configured.
-<img src="./assets/screenshot-skills.png" alt="Skill selection" width="700">
--->
+**2. Customize skills.** Add or remove from the interactive grid.
 
-After picking a stack, you can add or remove individual skills. 87+ skills are available across these categories:
+**3. Select subagents.** Choose which role-based subagents to include.
 
-**Web**<br>
-`React` `Vue` `Angular` `SolidJS` `Next.js` `Remix` `Nuxt` `SCSS Modules` `CVA` `Zustand` `Pinia` `NgRx SignalStore` `Jotai` `React Query` `SWR` `tRPC` `GraphQL` `React Hook Form` `Zod` `shadcn/ui` `Radix UI` `TanStack Table` `Vitest` `Playwright` `Cypress` `MSW` `Framer Motion` `Storybook` `Accessibility`
+**4. Compile.** After init, use `agentsinc edit` to change selections and `agentsinc compile` to rebuild.
 
-**API**<br>
-`Hono` `Express` `Fastify` `Drizzle` `Prisma` `Better Auth` `PostHog` `Resend` `Axiom + Pino + Sentry` `GitHub Actions`
+## Skills
 
-**Mobile**<br>
-`React Native` `Expo`
+87+ skills organized by category:
 
-**CLI**<br>
-`Commander` `oclif + Ink`
+**Web:** `React` `Vue` `Angular` `SolidJS` `Next.js` `Remix` `Nuxt` `SCSS Modules` `CVA` `Zustand` `Pinia` `NgRx SignalStore` `Jotai` `React Query` `SWR` `tRPC` `GraphQL` `React Hook Form` `Zod` `shadcn/ui` `Radix UI` `TanStack Table` `Vitest` `Playwright` `Cypress` `MSW` `Framer Motion` `Storybook` `Accessibility`
 
-**Infra**<br>
-`Turborepo` `Tooling` `Env config`
+**API:** `Hono` `Express` `Fastify` `Drizzle` `Prisma` `Better Auth` `PostHog` `Resend` `Axiom + Pino + Sentry` `GitHub Actions`
 
-**Meta**<br>
-`Code reviewing` `Research methodology` `Investigation requirements` `Anti-over-engineering` `Context management`
+**Mobile:** `React Native` `Expo`
 
-Skills live in the [Agents Inc marketplace](https://github.com/agents-inc/skills). Each one is a structured package with metadata, versioning, and compatibility declarations. Everything is validated against schemas, so if a skill name doesn't match or a config is malformed, you get an actual error instead of silent failure.
+**CLI:** `Commander` `oclif + Ink`
 
-### 3. Configure subagents
+**Infra:** `Turborepo` `Tooling` `Env config`
 
-<!-- TODO: Screenshot of the subagent configuration screen.
-Shows which agents will be compiled (web-developer, web-reviewer, web-tester, etc.)
-and which skills are mapped to each one.
-<img src="./assets/screenshot-agents.png" alt="Subagent configuration" width="700">
--->
+**Meta:** `Code reviewing` `Research methodology` `Investigation requirements` `Anti-over-engineering` `Context management`
 
-Each stack includes agents like `web-developer`, `api-developer`, `web-reviewer`, `web-tester`, `web-researcher`, `pattern-scout`, and `documentor`: roles that use the right skills for the job.
+## Subagents
 
----
+17 roles across 7 categories:
 
-Both modes (stack or from scratch) compile agents and generate a config at `.claude-src/config.yaml`. Use `agentsinc edit` to change your setup at any time.
+| Category         | Subagents                                                          |
+| ---------------- | ------------------------------------------------------------------ |
+| Developers       | `web-developer` `api-developer` `cli-developer` `web-architecture` |
+| Reviewers        | `web-reviewer` `api-reviewer` `cli-reviewer`                       |
+| Testers          | `web-tester` `cli-tester`                                          |
+| Researchers      | `web-researcher` `api-researcher`                                  |
+| Planning         | `web-pm`                                                           |
+| Pattern Analysis | `pattern-scout` `web-pattern-critique`                             |
+| Documentation    | `documentor`                                                       |
+| Meta             | `skill-summoner` `agent-summoner`                                  |
 
 ## Commands
 
-| Command        | Description                                                |
-| -------------- | ---------------------------------------------------------- |
-| `init`         | Interactive setup: pick skills, configure agents           |
-| `edit`         | Modify your skill selection                                |
-| `compile`      | Recompile agents after changes                             |
-| `update`       | Pull latest skills from source                             |
-| `search`       | Search skills across sources (interactive or static query) |
-| `eject`        | Export skills or agent partials for customization          |
-| `new skill`    | Scaffold a custom skill                                    |
-| `new agent`    | Scaffold a custom agent                                    |
-| `import skill` | Import a skill from an external GitHub repository          |
-| `doctor`       | Diagnose setup issues                                      |
-| `uninstall`    | Remove Agents Inc from your project                        |
+### Primary
 
-Every command supports `--dry-run` and `--source` flags. Run `agentsinc --help` for the full command list including `config`, `version`, `build`, `diff`, `outdated`, `validate`, `info`, and `list`.
+| Command   | Description                                                                 |
+| --------- | --------------------------------------------------------------------------- |
+| `init`    | Interactive setup wizard: pick a stack, customize skills, compile subagents |
+| `edit`    | Modify skill selection in the interactive wizard                            |
+| `compile` | Recompile subagents after changes                                           |
+| `update`  | Pull latest skills from source                                              |
 
-## Importing third-party skills
+### Customization
 
-Import skills from any GitHub repository into your local setup:
+| Command        | Description                                                                                                       |
+| -------------- | ----------------------------------------------------------------------------------------------------------------- |
+| `eject <type>` | Export for customization (`agent-partials`, `templates`, `skills`, `all`)                                         |
+| `new skill`    | Scaffold a custom skill ![Coming soon](https://img.shields.io/badge/coming%20soon-grey)                           |
+| `new agent`    | Scaffold a custom agent ![Coming soon](https://img.shields.io/badge/coming%20soon-grey)                           |
+| `import skill` | Import a skill from an external GitHub repository ![Coming soon](https://img.shields.io/badge/coming%20soon-grey) |
+| `search`       | Search skills across sources ![Coming soon](https://img.shields.io/badge/coming%20soon-grey)                      |
 
-```bash
-# List available skills from a repository
-agentsinc import skill github:your-org/skills --list
+### Diagnostics
 
-# Import a specific skill
-agentsinc import skill github:your-org/skills --skill react-best-practices
-
-# Import all skills from a repository
-agentsinc import skill github:your-org/skills --all
-```
+| Command     | Description                                                                                                  |
+| ----------- | ------------------------------------------------------------------------------------------------------------ |
+| `doctor`    | Diagnose setup issues                                                                                        |
+| `diff`      | Show changes between local and source skills ![Coming soon](https://img.shields.io/badge/coming%20soon-grey) |
+| `outdated`  | Check for skill updates                                                                                      |
+| `validate`  | Validate config and skill structure                                                                          |
+| `info`      | Show project configuration details                                                                           |
+| `uninstall` | Remove Agents Inc from your project                                                                          |
 
-Imported skills are copied to `.claude/skills/` and tracked with metadata for future updates.
+Run `agentsinc --help` for full usage.
 
 ## Customization
 
-Everything is configured through `.claude-src/config.yaml` with schema validation. There are progressive levels of customization depending on how deep you want to go:
+Progressive layers from config to full extension:
 
-**Edit the config directly** to change skill-to-agent mappings:
+| Layer           | How                                                                                            |
+| --------------- | ---------------------------------------------------------------------------------------------- |
+| Edit config     | `.claude-src/config.yaml`; skill-to-subagent mappings, preloaded/dynamic                       |
+| Wizard          | `agentsinc edit`; add/remove skills interactively                                              |
+| Eject partials  | `agentsinc eject agent-partials`; customize agent partials such as intro, workflow, and output |
+| Eject templates | `agentsinc eject templates`; modify Liquid templates                                           |
+| Eject skills    | `agentsinc eject skills`; fork skills for local editing                                        |
+| Custom skills   | `agentsinc new skill`; scaffold with proper metadata                                           |
+| Custom agents   | `agentsinc new agent`; scaffold agent files                                                    |
+| Custom sources  | Point to a private repo or local directory as a skill source                                   |
+| Plugins         | Package skills/subagents as Claude Code plugins for distribution                               |
 
-```yaml
-web-developer:
-  web-framework:
-    id: web-framework-react
-    preloaded: true
-  web-styling: web-styling-tailwind-v3
-```
+Templates apply globally across all subagents; partials apply to specific roles. Put shared conventions in a template for consistency, and role-specific behaviour in partials.
 
-**Eject** for deeper control:
+Import skills from any GitHub repository:
 
 ```bash
-# Eject agent partials (intro, workflow, critical requirements, etc.)
-npx @agents-inc/cli eject agent-partials
-
-# Eject the Liquid templates that control how agents are compiled
-npx @agents-inc/cli eject agent-partials --templates
-
-# Eject skills for local editing
-npx @agents-inc/cli eject skills
+agentsinc import skill github:your-org/skills --list
+agentsinc import skill github:your-org/skills --skill react-best-practices
+agentsinc import skill github:your-org/skills --all
 ```
 
-**Create custom skills** when you need something that doesn't exist:
+## Tech Stack
 
-```bash
-npx @agents-inc/cli new skill my-custom-skill
-```
+| Component    | Technology                            |
+| ------------ | ------------------------------------- |
+| Commands     | oclif                                 |
+| Terminal UI  | Ink + React                           |
+| Wizard state | Zustand                               |
+| Validation   | Zod (30+ schemas at parse boundaries) |
+| Compilation  | LiquidJS                              |
+| Testing      | Vitest                                |
 
-## Architecture
-
-The CLI uses oclif for commands, Ink (React) for the terminal UI, and Zustand for wizard state. The codebase is organized into domain-driven library modules with barrel exports.
+```
+src/cli/
+  commands/       # oclif command definitions
+  components/     # Ink/React terminal UI components
+  lib/            # Core logic (compiler, loader, resolver, schemas)
+  stores/         # Zustand state management
+  utils/          # Shared utilities
+src/agents/       # Subagent definitions (18 roles)
+config/           # Skills matrix, stacks, default mappings
+```
 
 See [docs/reference/architecture.md](./docs/reference/architecture.md) for the full reference.
 
 ## Links
 
-- [Agents Inc Skills](https://github.com/agents-inc/skills): the marketplace
+- [Plugin Marketplace](https://github.com/agents-inc/skills): browse and discover skills
+- [Architecture Reference](./docs/reference/architecture.md): full system documentation
 
 ## License