@namch/agent-assistant 1.3.0 → 1.3.2

package/documents/knowledge-overview/01-project-identity.md

@@ -0,0 +1,162 @@
+# Agent Assistant — Project Identity
+
+> **Purpose**: Defines the project's name, version, author, license, purpose, problem/solution, key benefits, and onboarding path
+> **Parent**: [00-index.md](./00-index.md)
+> **Last Updated**: 2026-03-26
+> **Generated By**: docs-core skill
+
+---
+
+## Basic Information
+
+| Field | Value |
+|-------|-------|
+| Name | Agent Assistant |
+| Package | `@namch/agent-assistant` |
+| Version | 1.3.0 |
+| Type | Multi-agent orchestration framework for AI coding assistants |
+| Author | NamCH |
+| License | MIT |
+| Repository | https://github.com/hainamchung/agent-assistant |
+| npm Registry | https://www.npmjs.com/package/@namch/agent-assistant |
+| Website | https://agent-assistant-ten.vercel.app |
+| Scope | `@namch` (public) |
+| Module System | CommonJS (CLI), ESM (web) |
+| Node.js Requirement | >=18.0.0 |
+
+---
+
+## Purpose
+
+Agent Assistant transforms a single AI coding assistant into a coordinated team of 21 specialist agents. It provides structured workflows, enforced quality gates, and automatic skill injection so that AI-assisted development follows consistent, high-quality patterns across every project and every supported platform.
+
+---
+
+## Problem Statement
+
+AI coding assistants (Cursor, Copilot, Claude Code, etc.) are powerful but operate as generalists by default. Without structure, they produce inconsistent outputs, miss quality checks, and require extensive manual prompt engineering for each task. Developers waste time re-configuring AI behavior per project and per tool, leading to:
+
+- **Inconsistent quality** — No enforced review, testing, or security checks
+- **Repeated setup** — Configuration lost between projects and tools
+- **Token waste** — Unfocused prompts produce verbose, often incorrect outputs
+- **No specialization** — A single agent tries to be backend engineer, tester, and reviewer simultaneously
+- **Platform lock-in** — Workflows built for one AI tool don't transfer to another
+
+---
+
+## Solution
+
+Agent Assistant solves these by:
+
+1. **Global one-time installation** — Markdown and YAML configurations installed into each AI tool's global directory (`~/.cursor/`, `~/.claude/`, `~/.copilot/`, etc.) apply to all projects automatically.
+2. **Orchestrator pattern** — The AI reads a platform entry file that transforms it into an Orchestrator role. It delegates work to specialist agents rather than attempting everything itself.
+3. **21 specialist agents** — Each with a defined profile, responsibility scope, and skill set. Examples: `backend-engineer`, `tester`, `security-engineer`, `tech-lead`.
+4. **14 structured commands** — `/cook`, `/code`, `/fix`, `/debug`, `/test`, `/plan`, `/design`, `/review`, `/report`, `/brainstorm`, `/docs`, `/deploy`, `/ask`, `/auto` — each with variant strategies (fast, hard, team).
+5. **Automatic skill injection (HSOL)** — 1,430+ skills across 19 domains resolved by agent profile and request context, with no manual skill selection required.
+6. **Cross-platform parity** — The same workflows and agent definitions work across all 5 supported AI coding tools.
+
+---
+
+## Key Benefits
+
+| Benefit | Detail |
+|---------|--------|
+| **70% faster delivery** | Structured workflows with parallel sub-agent execution eliminate manual coordination overhead |
+| **70% fewer bugs** | Quality gates in `:hard` variants enforce testing, review, and security checks before delivery |
+| **85% token savings** | Focused agent profiles and skill injection reduce prompt size and improve response relevance |
+| **Zero per-project config** | Global installation means every new project inherits the full framework automatically |
+| **Platform freedom** | Switch between Cursor, Copilot, Claude Code, Codex, or Antigravity without rewriting workflows |
+| **Built-in governance** | 10 Orchestration Laws (Single Point of Truth, Requirement Integrity, etc.) enforce consistent behavior |
+| **Error self-healing** | Classified recovery system (E1–E4) prevents silent failures and ensures transparent error handling |
+
+---
+
+## Target Users
+
+- **Individual developers** using AI coding assistants who want structured, repeatable workflows
+- **Teams** wanting consistent AI-assisted development practices across members and tools
+- **Organizations** standardizing on AI-augmented development with quality governance
+
+---
+
+## Project Classification
+
+| Attribute | Value |
+|-----------|-------|
+| Architecture | Plugin-based Orchestrator Framework |
+| Delivery Model | CLI tool + Markdown/YAML configurations |
+| Runtime Dependencies | None (zero production deps) |
+| State Management | Stateless — no database, no server, no persistent runtime |
+| Distribution | npm public registry + GitHub source |
+| Versioning | Semantic Versioning via semantic-release |
+| Commit Convention | Conventional Commits (v7.0.2) |
+
+---
+
+## First 60 Minutes — Onboarding Checklist
+
+This checklist takes a new developer or contributor from zero to productive with Agent Assistant in approximately one hour.
+
+### Minutes 0–10: Environment Setup
+
+- [ ] Verify Node.js >=18.0.0 is installed: `node --version`
+- [ ] Verify npm is available: `npm --version`
+- [ ] Verify git is available: `git --version`
+- [ ] Ensure at least one supported AI tool is installed (Cursor, VS Code with Copilot, Claude Code CLI, Codex, or Antigravity)
+
+### Minutes 10–20: Install and Verify
+
+- [ ] Install globally: `npm install -g @namch/agent-assistant@latest`
+- [ ] Install for your AI tool: `agent-assistant install --all` (or specify one: `agent-assistant install cursor`)
+- [ ] Verify installation: `agent-assistant list`
+- [ ] Confirm files exist in the tool's directory (e.g., `ls ~/.cursor/skills/agent-assistant/` for Cursor)
+
+### Minutes 20–30: Read Core Concepts
+
+- [ ] Read the README: [../../README.md](../../README.md) — understand the value prop, commands, and agents
+- [ ] Read Core Rules: [../../rules/CORE.md](../../rules/CORE.md) — understand the Orchestrator identity and 10 Laws
+- [ ] Browse agent definitions: [../../agents/](../../agents/) — open 2–3 agent files to see their structure (e.g., `backend-engineer.md`, `tester.md`)
+
+### Minutes 30–40: Run Your First Commands
+
+- [ ] Open a project in your AI tool (Cursor, VS Code, or Claude Code CLI)
+- [ ] Try a simple build: `/cook:fast "add a hello world endpoint"`
+- [ ] Try a planning command: `/plan "build a user authentication system"`
+- [ ] Try an ask command: `/ask "explain the project structure"`
+- [ ] Observe how the Orchestrator delegates to specialist agents
+
+### Minutes 40–50: Explore the Framework
+
+- [ ] Browse the commands directory: [../../commands/](../../commands/) — see how commands route to variants
+- [ ] Open one command variant (e.g., `commands/cook/fast.md`) to understand workflow phases
+- [ ] Browse `matrix-skills/` to see how skills are organized: [../../matrix-skills/_index.yaml](../../matrix-skills/_index.yaml)
+- [ ] Check `skills/` to see the scale of domain knowledge available
+
+### Minutes 50–60: Generate Project Documentation
+
+- [ ] Run `/docs:core` in your AI tool to generate technical documentation for your current project
+- [ ] Verify that `./documents/` was created with knowledge files
+- [ ] Try `/docs:business` for business requirements documentation
+- [ ] Review the generated docs and understand how agents used your project context
+
+### Post-Onboarding Next Steps
+
+- [ ] Try `:hard` variant for a complex feature: `/cook:hard "implement OAuth 2.0 flow"`
+- [ ] Try `:team` variant for adversarial collaboration: `/cook:team "implement payment processing"`
+- [ ] Read [../../rules/PHASES.md](../../rules/PHASES.md) for understanding workflow phases
+- [ ] Read [../../rules/AGENTS.md](../../rules/AGENTS.md) for tiered execution details
+- [ ] If contributing: read the Contributing section in the README and follow fork → branch → PR workflow
+
+---
+
+## Evidence Sources
+
+| Source | Path |
+|--------|------|
+| Package manifest (name, version, author, license, engines) | [../../package.json](../../package.json) |
+| README (purpose, benefits, metrics, installation) | [../../README.md](../../README.md) |
+| Core rules (Orchestrator identity, 10 Laws) | [../../rules/CORE.md](../../rules/CORE.md) |
+| Agent definitions (21 agents) | [../../agents/](../../agents/) |
+| CLI installer (platform paths, tool configs) | [../../cli/install.js](../../cli/install.js) |
+| Commands directory (14 commands + variants) | [../../commands/](../../commands/) |
+| Knowledge source base index | [../knowledge-source-base/00-index.md](../knowledge-source-base/00-index.md) |

package/documents/knowledge-overview/02-tech-stack.md

@@ -0,0 +1,119 @@
+# Agent Assistant — Technology Stack
+
+> **Purpose**: Categorized inventory of all technologies, frameworks, and tools used in the project with versions and evidence
+> **Parent**: [00-index.md](./00-index.md)
+> **Last Updated**: 2026-03-26
+> **Generated By**: docs-core skill
+
+---
+
+## Stack Overview
+
+Agent Assistant is split into two distinct codebases under one repository:
+
+1. **CLI Framework** (core product) — Pure Node.js with zero dependencies, CommonJS modules
+2. **Marketing Website** (`web/`) — React 19 + Vite + Tailwind CSS v4 SPA deployed to Vercel
+
+The CLI has no production dependencies. All npm packages listed in `package.json` are devDependencies for release automation and git hooks.
+
+---
+
+## Runtime
+
+| Technology | Version | Role | Evidence |
+|-----------|---------|------|----------|
+| Node.js | >=18.0.0 | Runtime for CLI installer and test runner | [../../package.json](../../package.json) — `engines.node` |
+| npm | (bundled with Node.js) | Package manager, global install target | [../../package.json](../../package.json) — `publishConfig`, `bin` |
+
+---
+
+## CLI / Core Framework
+
+| Technology | Version | Role | Evidence |
+|-----------|---------|------|----------|
+| Vanilla JavaScript (CommonJS) | ES2022+ | CLI installer logic — file copying, path resolution, progress display | [../../cli/install.js](../../cli/install.js) |
+| Node.js built-in: `fs` | N/A | File system operations — read, write, copy, mkdir, stat | [../../cli/install.js](../../cli/install.js) |
+| Node.js built-in: `path` | N/A | Cross-platform path resolution | [../../cli/install.js](../../cli/install.js) |
+| Node.js built-in: `os` | N/A | Home directory detection (`os.homedir()`) | [../../cli/install.js](../../cli/install.js) |
+| Node.js built-in: `readline` | N/A | Terminal progress bar rendering | [../../cli/install.js](../../cli/install.js) |
+| Node.js built-in: `node:test` | N/A | Test runner via `node --test` | [../../package.json](../../package.json) — `scripts.test` |
+| Markdown | N/A | Agent definitions, commands, rules, skill docs | [../../agents/](../../agents/), [../../commands/](../../commands/), [../../rules/](../../rules/) |
+| YAML | N/A | Matrix skill registries (19 domain files) | [../../matrix-skills/](../../matrix-skills/) |
+
+---
+
+## Web Frontend (Marketing Site)
+
+| Technology | Version | Role | Evidence |
+|-----------|---------|------|----------|
+| React | ^19.2.0 | UI component framework | [../../web/package.json](../../web/package.json) |
+| Vite | devDep | Build tool and dev server | [../../web/package.json](../../web/package.json) |
+| Tailwind CSS | ^4.1.18 (v4) | Utility-first CSS framework | [../../web/package.json](../../web/package.json) |
+| Framer Motion | ^12.29.2 | Animation library for page transitions and UI effects | [../../web/package.json](../../web/package.json) |
+| react-router-dom | ^7.13.0 | Client-side routing | [../../web/package.json](../../web/package.json) |
+| @xyflow/react | ^12.10.0 | Interactive node/flow diagrams (architecture visualization) | [../../web/package.json](../../web/package.json) |
+| lucide-react | ^0.563.0 | Icon library | [../../web/package.json](../../web/package.json) |
+
+---
+
+## CI/CD and Release
+
+| Technology | Version | Role | Evidence |
+|-----------|---------|------|----------|
+| semantic-release | ^22.0.12 | Automated versioning and npm publishing | [../../package.json](../../package.json) — `devDependencies` |
+| @semantic-release/commit-analyzer | ^11.1.0 | Determines version bump from commit messages | [../../package.json](../../package.json) — `devDependencies` |
+| @semantic-release/release-notes-generator | ^12.1.0 | Generates release notes from commits | [../../package.json](../../package.json) — `devDependencies` |
+| @semantic-release/changelog | ^6.0.3 | Updates CHANGELOG.md on release | [../../package.json](../../package.json) — `devDependencies` |
+| @semantic-release/npm | ^11.0.2 | Publishes to npm registry | [../../package.json](../../package.json) — `devDependencies` |
+| @semantic-release/git | ^10.0.1 | Commits version bumps and changelog back to repo | [../../package.json](../../package.json) — `devDependencies` |
+| @semantic-release/github | ^9.2.6 | Creates GitHub releases | [../../package.json](../../package.json) — `devDependencies` |
+| @semantic-release/exec | ^6.0.3 | Runs custom commands during release lifecycle | [../../package.json](../../package.json) — `devDependencies` |
+| conventional-changelog-conventionalcommits | ^7.0.2 | Commit message parsing preset | [../../package.json](../../package.json) — `devDependencies` |
+
+---
+
+## Git and Code Quality
+
+| Technology | Version | Role | Evidence |
+|-----------|---------|------|----------|
+| Husky | ^8.0.3 | Git hooks manager (pre-commit, commit-msg) | [../../package.json](../../package.json) — `devDependencies`, `scripts.prepare` |
+| Conventional Commits | v7.0.2 (via preset) | Commit message format standard (`feat:`, `fix:`, `docs:`, etc.) | [../../package.json](../../package.json) — `conventional-changelog-conventionalcommits` |
+| Git | N/A | Version control | Repository root |
+
+---
+
+## Hosting and Deployment
+
+| Technology | Version | Role | Evidence |
+|-----------|---------|------|----------|
+| npm Registry | N/A | Package distribution (public scoped: `@namch`) | [../../package.json](../../package.json) — `publishConfig.access: "public"` |
+| GitHub | N/A | Source code hosting, issue tracking, releases | [../../README.md](../../README.md) — repository links |
+| Vercel | N/A | Marketing website hosting | [../../README.md](../../README.md) — website URL `agent-assistant-ten.vercel.app` |
+
+---
+
+## What Is NOT in the Stack
+
+| Category | Status | Notes |
+|----------|--------|-------|
+| Database | None | Stateless framework — no persistent storage |
+| ORM | None | No database means no ORM |
+| Backend Server | None | CLI-only, no HTTP server in the core product |
+| Authentication | None | No user accounts or authentication system |
+| Containerization | None | No Docker or container configs |
+| Testing Framework | Built-in only | Uses `node --test` (Node.js native test runner), no Jest/Mocha/Vitest |
+| Linting | Manual only | `scripts.lint` runs `node --check` for syntax validation; markdownlint suggested but not enforced |
+| Type System | None | Plain JavaScript, no TypeScript |
+
+---
+
+## Evidence Sources
+
+| Source | Path |
+|--------|------|
+| Root package manifest (all devDeps, engines, scripts) | [../../package.json](../../package.json) |
+| CLI installer (Node.js builtins used) | [../../cli/install.js](../../cli/install.js) |
+| README (hosting, platform references) | [../../README.md](../../README.md) |
+| Web directory (frontend stack) | [../../web/](../../web/) |
+| Matrix skills (YAML usage) | [../../matrix-skills/](../../matrix-skills/) |
+| Agent definitions (Markdown usage) | [../../agents/](../../agents/) |

package/documents/knowledge-overview/03-features.md

@@ -0,0 +1,232 @@
+# Agent Assistant — Features
+
+> **Purpose**: Comprehensive inventory of key features, capabilities, performance metrics, and component counts
+> **Parent**: [00-index.md](./00-index.md)
+> **Last Updated**: 2026-03-26
+> **Generated By**: docs-core skill
+
+---
+
+## Performance Metrics
+
+| Metric | Improvement | Context |
+|--------|------------|---------|
+| Time-to-Production | 70% faster | Structured workflows with parallel sub-agent execution eliminate manual coordination |
+| Bug Rate | 70% reduction | Quality gates in `:hard` variants enforce testing, review, and security before delivery |
+| Token Cost | 85% savings | Focused agent profiles and HSOL skill injection reduce prompt size and improve relevance |
+
+---
+
+## Key Features
+
+### 1. One-Time Setup, Forever Use
+
+Install once at the global level (`~/.cursor/`, `~/.claude/`, `~/.copilot/`, `~/.codex/`, `~/.gemini/`) and the framework auto-applies to every project opened in that AI tool. No per-project configuration, no repeated setup, no boilerplate.
+
+**How it works**: The CLI installer (`cli/install.js`) copies all agents, commands, rules, skills, and matrix-skills into the AI tool's global configuration directory. The tool reads these files on every session start.
+
+**Evidence**: [../../cli/install.js](../../cli/install.js), [../../README.md](../../README.md)
+
+---
+
+### 2. Sub-Agent Orchestration
+
+When the platform supports it (Claude Code with `runSubagent`, Cursor Max mode), the Orchestrator spawns specialized sub-agents that execute tasks in parallel. A backend engineer, tester, and reviewer can work simultaneously on different aspects of the same feature.
+
+**How it works**: The Orchestrator follows Tiered Execution — TIER 1 uses `runSubagent` for isolated parallel execution; TIER 2 falls back to embodying the agent in-context when sub-agents are unavailable.
+
+**Evidence**: [../../rules/CORE.md](../../rules/CORE.md), [../../rules/AGENTS.md](../../rules/AGENTS.md)
+
+---
+
+### 3. Multi-Platform Support
+
+The same agent definitions, commands, and workflows work across 5 AI coding tools. Write once, use everywhere.
+
+| Platform | Install Path | Status |
+|----------|-------------|--------|
+| Cursor | `~/.cursor/skills/agent-assistant/` | Full support |
+| GitHub Copilot | `~/.copilot/skills/agent-assistant/` | Full support |
+| Claude Code | `~/.claude/skills/agent-assistant/` | Full support |
+| Codex | `~/.codex/skills/agent-assistant/` | Full support |
+| Antigravity/Gemini | `~/.gemini/antigravity/skills/agent-assistant/` | Full support |
+
+**How it works**: The CLI installer has platform-specific path configurations and template variable replacements (`{TOOL}`, `{HOME}`) that adapt the same source files for each platform's directory structure.
+
+**Evidence**: [../../cli/install.js](../../cli/install.js) — `TOOLS` configuration object
+
+---
+
+### 4. Matrix Skill Discovery (HSOL)
+
+The Hybrid Skill Orchestration Layer automatically resolves and injects relevant skills based on the agent's profile and the task context. No manual skill selection required.
+
+**Scale**: 1,430+ skill modules across 19 domains (ai-ml, architecture, backend, cloud, data, design, devops, frontend, gaming, languages, management, mcp, mobile, performance, planning, quality, research, security, tools).
+
+**Resolution algorithm**:
+1. Parse agent profile from frontmatter
+2. Load inherited domains from `matrix-skills/_index.yaml`
+3. Filter skills by relevance mapping
+4. Apply priority thresholds (critical >=9, core >=7, minimum >=5)
+5. Calculate fitness scores (semantic match 35%, specificity 25%, trust 20%, freshness 10%, success rate 10%)
+6. Return sorted skill set
+
+**Dynamic discovery**: When matrix fitness score falls below 0.8 for `:hard` or `:team` variants, the `find-skills` mechanism can discover additional community skills at runtime.
+
+**Evidence**: [../../rules/SKILLS.md](../../rules/SKILLS.md), [../../matrix-skills/_index.yaml](../../matrix-skills/_index.yaml), [../../matrix-skills/](../../matrix-skills/)
+
+---
+
+### 5. 21 Specialist Agents
+
+Each agent is a Markdown file with a defined role, profile, responsibilities, and skill requirements.
+
+| Domain | Agents |
+|--------|--------|
+| **Implementation** | backend-engineer, frontend-engineer, mobile-engineer, game-engineer |
+| **Architecture** | tech-lead, database-architect |
+| **Quality** | tester, reviewer, debugger, security-engineer |
+| **Planning** | planner, brainstormer, business-analyst |
+| **Support** | designer, devops-engineer, docs-manager, performance-engineer, researcher, scouter, project-manager, reporter |
+
+**Team configurations**: 17 teams with 3 agents each (51 team agents total), enabling structured adversarial collaboration via the `:team` variant.
+
+| Team | Purpose |
+|------|---------|
+| backend-team | Backend implementation with review |
+| frontend-team | Frontend implementation with review |
+| fullstack-team | Full-stack feature development |
+| mobile-team | Mobile app development |
+| game-team | Game development |
+| database-team | Database design and migration |
+| qa-team | Quality assurance and testing |
+| security-team | Security audit and hardening |
+| devops-team | Infrastructure and deployment |
+| design-team | UI/UX design |
+| planning-team | Project planning and strategy |
+| research-team | Technical research |
+| performance-team | Performance optimization |
+| docs-team | Documentation |
+| debug-team | Debugging and troubleshooting |
+| report-team | Reporting and analysis |
+| project-team | Project management |
+
+**Evidence**: [../../agents/](../../agents/), [../../agents/teams/](../../agents/teams/)
+
+---
+
+### 6. 14 Structured Commands
+
+Commands are workflow definitions that orchestrate agent collaboration across phases.
+
+| Category | Commands | Description |
+|----------|----------|-------------|
+| **Build** | `/cook`, `/code`, `/fix` | Feature implementation, code generation, bug fixing |
+| **Quality** | `/test`, `/review`, `/debug` | Test generation, code review, debugging |
+| **Plan** | `/plan`, `/brainstorm`, `/design` | Implementation planning, idea exploration, UI/UX design |
+| **Docs** | `/docs` | Documentation generation (core, business, audit variants) |
+| **Report** | `/report` | Status reports, sprint summaries |
+| **Deploy** | `/deploy` | Deployment checks, preview, production release |
+| **Inquiry** | `/ask` | Q&A, explanations, research |
+| **Auto** | `/auto` | Autonomous task execution |
+
+**Evidence**: [../../commands/](../../commands/)
+
+---
+
+### 7. 4 Variant Strategies
+
+Each command supports up to 4 variants that control scope, agent count, and quality gates:
+
+| Variant | Scope | Agents | Quality Gates | Use Case |
+|---------|-------|--------|---------------|----------|
+| `:fast` | Minimal | 2–3 | Basic | Simple tasks, quick fixes |
+| `:hard` | Full | 5–8 | Comprehensive (test, review, security) | Complex features, production code |
+| `:team` | Adversarial | 3 per team (Golden Triangle) | Debate mechanism | Critical features needing multiple perspectives |
+
+**Evidence**: [../../README.md](../../README.md), [../../commands/cook/](../../commands/cook/)
+
+---
+
+### 8. Golden Triangle Team Protocol
+
+The `:team` variant uses an adversarial collaboration model with three roles:
+
+1. **Tech Lead** — Designs the approach and makes architectural decisions
+2. **Executor** — Implements the solution (backend-engineer, frontend-engineer, etc.)
+3. **Reviewer** — Challenges the implementation, finds flaws, ensures quality
+
+A built-in debate mechanism allows the Reviewer to push back on the Executor's work, forcing improvements before acceptance.
+
+**Evidence**: [../../rules/TEAMS.md](../../rules/TEAMS.md), [../../agents/teams/](../../agents/teams/)
+
+---
+
+### 9. 10 Orchestration Laws
+
+Formal governance rules that constrain Orchestrator behavior:
+
+1. **Single Point of Truth** — One authoritative source for each type of information
+2. **Requirement Integrity** — User requirements preserved without modification
+3. **Phase Completion** — Each phase must meet exit criteria before proceeding
+4. **Agent Specialization** — Agents only perform tasks within their defined scope
+5. **Transparent Delegation** — All delegations are visible and traceable
+6. **No Silent Failures** — Every error must be surfaced and handled
+7. **Skill Authority** — HSOL-resolved skills override generic knowledge
+8. **Output Verification** — Deliverables verified against acceptance criteria
+9. **Context Preservation** — Project context maintained across agent handoffs
+10. **Minimal Intervention** — Orchestrator delegates; never implements directly
+
+**Evidence**: [../../rules/CORE.md](../../rules/CORE.md)
+
+---
+
+### 10. Error Self-Healing (E1–E4)
+
+Classified error recovery system that prevents silent failures:
+
+| Class | Severity | Recovery |
+|-------|----------|----------|
+| E1 | Warning | Auto-retry with adjusted parameters |
+| E2 | Recoverable | Fallback strategy (TIER 2 embodiment) |
+| E3 | Blocking | Pause and request user clarification |
+| E4 | Critical | Halt with full diagnostic report |
+
+**Evidence**: [../../rules/ERRORS.md](../../rules/ERRORS.md)
+
+---
+
+## Capability Summary
+
+| Capability | Count |
+|-----------|-------|
+| Specialist Agents | 21 |
+| Team Configurations | 17 (51 team agents) |
+| Commands | 14 |
+| Variants | 4 (fast, hard, team) |
+| Skill Modules | 1,430+ |
+| Skill Domains | 19 |
+| Orchestration Rules | 7 files |
+| Orchestration Laws | 10 |
+| Supported Platforms | 5 |
+| Error Recovery Classes | 4 (E1–E4) |
+| Matrix Skill Registries | 19 YAML files + 2 meta files |
+
+---
+
+## Evidence Sources
+
+| Source | Path |
+|--------|------|
+| README (features, metrics, agents, commands) | [../../README.md](../../README.md) |
+| Core rules (10 Laws, Tiered Execution) | [../../rules/CORE.md](../../rules/CORE.md) |
+| Skills rules (HSOL algorithm) | [../../rules/SKILLS.md](../../rules/SKILLS.md) |
+| Teams rules (Golden Triangle) | [../../rules/TEAMS.md](../../rules/TEAMS.md) |
+| Errors rules (E1–E4 classification) | [../../rules/ERRORS.md](../../rules/ERRORS.md) |
+| Agents rules (Tiered Execution) | [../../rules/AGENTS.md](../../rules/AGENTS.md) |
+| Agent definitions (21 agents) | [../../agents/](../../agents/) |
+| Team definitions (17 teams) | [../../agents/teams/](../../agents/teams/) |
+| Commands (14 commands) | [../../commands/](../../commands/) |
+| Matrix skills (19 domains) | [../../matrix-skills/](../../matrix-skills/) |
+| Skills modules (1,430+) | [../../skills/](../../skills/) |
+| CLI installer (platform support) | [../../cli/install.js](../../cli/install.js) |