azrole 3.0.0

package/docs/case-studies/book-manuscript.md

@@ -0,0 +1,301 @@
+# AZROLE — Full Setup Log
+
+**Project**: Premium nonfiction biography (150K words, 24 chapters)
+**Date**: March 12, 2026
+**Starting Level**: 0 / 10 (Terminal Tourist)
+**Final Level**: 5 / 10 (Multi-Agent)
+
+---
+
+## Step 1: Installation (`npx azrole init`)
+
+Ran `npx azrole init` which installed the AZROLE package globally into `~/.claude/commands/`.
+
+### What it installed:
+
+| File | Location | Purpose |
+|------|----------|---------|
+| `dream.md` | `~/.claude/commands/` | Slash command — build a project from an idea |
+| `level-up.md` | `~/.claude/commands/` | Slash command — check level, build next one |
+| `evolve.md` | `~/.claude/commands/` | Slash command — auto-improve environment |
+| `fix.md` | `~/.claude/commands/` | Slash command — describe a bug, get it fixed |
+| `ship.md` | `~/.claude/commands/` | Slash command — commit and push changes |
+| `explain.md` | `~/.claude/commands/` | Slash command — explain code/errors |
+| `status.md` | `~/.claude/commands/` | Slash command — quick project health check |
+| `setup.md` | `~/.claude/commands/` | Slash command — scan existing project, build environment |
+| `orchestrator.md` | `~/.claude/agents/` | Master orchestrator agent (the brain behind all levels) |
+
+These are **global** — available in every project, not just this one.
+
+### What was skipped:
+- `orchestrator.md` and `dream.md` already existed from a prior install — skipped (would need `--force` to overwrite)
+
+---
+
+## Step 2: Project Scan (`/setup`)
+
+The `/setup` command triggered a full scan of the project directory.
+
+### What was scanned:
+
+1. **Root directory listing** — found 3 files:
+   - `manuscript.tex` (main LaTeX document)
+   - `docuthriller_memoir_premium_2026_premium.sty` (custom style package)
+   - `Chain_Continuity_Audit.md` (editorial audit report)
+
+2. **File content analysis**:
+   - Read `manuscript.tex` — identified LuaLaTeX + memoir class, EB Garamond fonts, custom commands (`\visiontwenty`, `\termdef`, `\concept`), 6×9 trade paperback specs
+   - Read the `.sty` file — confirmed engine requirement (LuaLaTeX only), all package options, font cascade logic
+   - Read the audit report — mapped all 24 chapters across 5 parts, identified known transition issues, extracted editorial standards
+
+3. **Existing Claude config** — found only `settings.local.json` with one permission rule. No CLAUDE.md, no MCP, no skills, no memory, no agents.
+
+4. **Additional working directories** — detected `azcomply-flow/frontend/src/core/i18n` and `azcomply-flow/frontend/src/messages` (separate project, not part of this setup)
+
+### Scan verdict:
+- **Project type**: Book manuscript (not a software project)
+- **Tech stack**: LaTeX/LuaLaTeX + memoir + custom .sty
+- **Current level**: 0 / 10 — nothing configured
+- **Target level**: 5 / 10 (Multi-Agent)
+
+---
+
+## Step 3: Directory Scaffolding
+
+Created the infrastructure directories:
+
+```
+.devteam/                    # AZROLE working directory
+.claude/agents/              # Specialized subagents
+.claude/commands/            # Slash commands (project-specific)
+.claude/skills/              # Technology skill files
+.claude/memory/              # Persistent memory system
+scripts/                     # Automation scripts (for future levels)
+```
+
+---
+
+## Step 4: Blueprint Generation
+
+Created `.devteam/blueprint.json` — the master plan that all subsequent levels reference.
+
+### Blueprint contents:
+- **Project metadata**: name, type (book-manuscript), author, description
+- **Tech stack**: LuaLaTeX engine, memoir class, .sty package, font list, build command
+- **Content structure**: 24 chapters, 5 parts, ~150K words, modular LaTeX input
+- **File map**: role of each file and its dependencies
+- **Planned agents**: dev-writer, dev-editor, dev-typesetter, dev-reviewer
+- **Planned skills**: latex-memoir, editorial-craft, project-conventions
+- **Planned commands**: build, audit-chapter, new-chapter, review
+
+---
+
+## Step 5: Level 1 — CLAUDE.md
+
+Created `CLAUDE.md` in the project root (73 lines).
+
+### What it contains:
+
+| Section | Content |
+|---------|---------|
+| Header | Project name, author, word count, target format |
+| Architecture | Engine (LuaLaTeX), class (memoir), style (.sty v1.3), build command, font cascade |
+| File Structure | Map of all 4 files with roles |
+| Manuscript Structure | Table of all 5 parts with chapter ranges and time periods |
+| Conventions | Custom LaTeX commands, typography rules, style options |
+| Editorial Standards | Execution lens framing, 5 thematic threads, cultural sensitivity rules |
+| Known Issues | 5 priority issues from the audit report (critical → polish) |
+| Agent Routing | Which agent handles which responsibility |
+| Available Commands | 4 slash commands with descriptions |
+
+**Purpose**: Any Claude conversation in this project now immediately understands the full context — what the project is, how it's structured, what the rules are, and who handles what.
+
+---
+
+## Step 6: Level 2 — MCP Configuration
+
+**Skipped intentionally.** This is a manuscript project with no external services (no databases, no APIs, no GitHub integration needed for the writing workflow). MCP servers would add complexity with no benefit.
+
+---
+
+## Step 7: Level 3 — Skills & Commands
+
+### Skills Created (3 files):
+
+#### 1. `.claude/skills/latex-memoir/SKILL.md`
+- **Triggers on**: LaTeX, LuaLaTeX, memoir, typography, fonts, compilation, .sty, page layout
+- **Contains**: Engine & build rules, style package option reference, custom command docs, typography rules, anti-patterns (what never to do)
+
+#### 2. `.claude/skills/editorial-craft/SKILL.md`
+- **Triggers on**: Writing, editing, reviewing, transitions, narrative voice, pacing, cultural sensitivity
+- **Contains**: Narrative voice guidelines, five thematic threads reference, transition standards, cultural sensitivity rules, chapter quality checklist
+
+#### 3. `.claude/skills/project-conventions/SKILL.md`
+- **Triggers on**: Creating or editing any file
+- **Contains**: File naming rules, LaTeX conventions, editorial conventions (chapter numbering, audit format), production specs
+
+### Commands Created (4 files):
+
+#### 1. `.claude/commands/build.md`
+- **Invoked with**: `/build`
+- **Does**: Runs `latexmk -lualatex manuscript.tex`, diagnoses errors from .log, fixes issues, re-compiles
+
+#### 2. `.claude/commands/audit-chapter.md`
+- **Invoked with**: `/audit-chapter [chapter number or name]`
+- **Does**: Reads chapter content + audit report + editorial standards → evaluates transitions, threads, chronology, sensitivity, evidence → rates each dimension → provides actionable recommendations
+
+#### 3. `.claude/commands/new-chapter.md`
+- **Invoked with**: `/new-chapter [description]`
+- **Does**: Determines placement in 5-part structure → generates scaffold with epigraph, hook, sections, handoff, source placeholders → adds to content file
+
+#### 4. `.claude/commands/review.md`
+- **Invoked with**: `/review`
+- **Does**: Full manuscript review — continuity, thematic coherence, chronology, cultural sensitivity, prose quality, production readiness → structured report with delta comparison to previous audit
+
+---
+
+## Step 8: Level 4 — Memory System
+
+### Files Created (4 files):
+
+#### 1. `.claude/memory/MEMORY.md` (master index — 32 lines)
+- Quick Context: what the project is in 2 sentences
+- Architecture snapshot: 3 key files and their roles
+- Critical Rules: placeholder (filled as lessons are learned)
+- Active Patterns: execution lens, thematic threads, cultural sensitivity
+- Known Gotchas: weak transitions, Part IV timeline overlap, LuaLaTeX requirement
+- Recent Decisions: placeholder
+- See Also: links to other memory files
+
+#### 2. `.claude/memory/codebase-map.md`
+- File-by-file index with roles and dependencies
+- Style package key exports
+- Custom command reference
+
+#### 3. `.claude/memory/decisions.md`
+- ADR-001: Why LuaLaTeX + memoir (OpenType fonts, professional quality)
+- ADR-002: Why single content file (simpler pipeline)
+
+#### 4. `.claude/memory/patterns.md`
+- Narrative patterns: execution lens, thematic threading, transition craftsmanship, temporal anchoring
+- LaTeX patterns: font cascade, conditional loading, option-driven style, strict penalties
+
+**Purpose**: Agents read MEMORY.md before starting any task. This gives them instant context without re-scanning the project every time. Patterns and gotchas prevent repeated mistakes.
+
+---
+
+## Step 9: Level 5 — Specialized Agents
+
+### Agents Created (4 files):
+
+#### 1. `.claude/agents/dev-writer.md` (model: sonnet)
+- **Role**: Chapter drafting, prose creation, narrative voice
+- **Owns**: All chapter content in `manuscript_content.tex`
+- **Protocol**: Reads MEMORY.md → editorial-craft skill → audit report → existing content → writes in project voice → reports which threads were advanced
+
+#### 2. `.claude/agents/dev-editor.md` (model: opus)
+- **Role**: Continuity, transitions, structural editing, thematic coherence
+- **Owns**: Structural integrity of the manuscript
+- **Protocol**: Reads audit report as primary reference → rates transitions ★–★★★★★ → tracks 5 threads → checks chronology → reports issues with priority ranking
+- **Has**: Pre-loaded list of known priority issues from the audit
+
+#### 3. `.claude/agents/dev-typesetter.md` (model: sonnet)
+- **Role**: LaTeX compilation, typography, layout, production
+- **Owns**: `manuscript.tex` and the `.sty` file
+- **Protocol**: Reads latex-memoir skill → compiles → diagnoses errors from .log → fixes → verifies output quality
+
+#### 4. `.claude/agents/dev-reviewer.md` (model: opus)
+- **Role**: Full quality audit, publication readiness, final sign-off
+- **Owns**: Quality assurance gate
+- **Protocol**: Reads everything → per-chapter assessment → manuscript-wide assessment → benchmarks against premium biography standards → publication readiness verdict
+
+### Model Routing Logic:
+- **Opus** (more capable, slower): Used for dev-editor and dev-reviewer — tasks requiring deep analysis, judgment, and quality assessment
+- **Sonnet** (fast, efficient): Used for dev-writer and dev-typesetter — tasks requiring generation and technical execution
+
+---
+
+## Step 10: Verification Scan
+
+Final environment scan confirmed all levels:
+
+```
+Level 1: CLAUDE.md               ✅ 73 lines
+Level 2: MCP Servers             ⊘  N/A (no external services)
+Level 3: Skills & Commands       ✅ 3 skills, 4 commands
+Level 4: Memory System           ✅ 32-line index + 4 memory files
+Level 5: Multi-Agent             ✅ 4 specialized agents
+```
+
+---
+
+## Final File Tree
+
+```
+book-project/
+├── CLAUDE.md                                        # Level 1: Project conventions
+├── manuscript.tex                                    # Main LaTeX document
+├── manuscript_content.tex                            # Chapter content (referenced)
+├── docuthriller_memoir_premium_2026_premium.sty     # Custom style package
+├── Chain_Continuity_Audit.md                         # Editorial audit report
+│
+├── .devteam/
+│   └── blueprint.json                               # Master project blueprint
+│
+├── .claude/
+│   ├── settings.local.json                          # Permissions
+│   │
+│   ├── commands/                                    # Level 3: Slash commands
+│   │   ├── build.md                                 #   /build
+│   │   ├── audit-chapter.md                         #   /audit-chapter
+│   │   ├── new-chapter.md                           #   /new-chapter
+│   │   └── review.md                                #   /review
+│   │
+│   ├── skills/                                      # Level 3: Technology skills
+│   │   ├── latex-memoir/SKILL.md                    #   LaTeX/LuaLaTeX patterns
+│   │   ├── editorial-craft/SKILL.md                 #   Narrative & editorial standards
+│   │   └── project-conventions/SKILL.md             #   Naming, structure, production
+│   │
+│   ├── memory/                                      # Level 4: Persistent memory
+│   │   ├── MEMORY.md                                #   Master index (32 lines)
+│   │   ├── codebase-map.md                          #   File index & dependencies
+│   │   ├── decisions.md                             #   Architecture decision records
+│   │   └── patterns.md                              #   Discovered patterns
+│   │
+│   └── agents/                                      # Level 5: Specialized agents
+│       ├── dev-writer.md                            #   Prose & chapter drafting (sonnet)
+│       ├── dev-editor.md                            #   Continuity & structure (opus)
+│       ├── dev-typesetter.md                        #   LaTeX & production (sonnet)
+│       ├── dev-reviewer.md                          #   Quality audit (opus)
+│       └── orchestrator.md                          #   AZROLE orchestrator (global)
+│
+└── scripts/                                         # (empty — for Level 6+)
+```
+
+---
+
+## What's Available Now
+
+| Command | What it does |
+|---------|-------------|
+| `/build` | Compile manuscript with LuaLaTeX |
+| `/audit-chapter [N]` | Editorial audit of a specific chapter |
+| `/new-chapter [desc]` | Scaffold a new chapter |
+| `/review` | Full manuscript review pass |
+| `/level-up` | Build the next level (Level 6: headless scripts & git hooks) |
+| `/evolve` | Auto-detect gaps and improve the environment |
+| `/status` | Quick project health check |
+
+---
+
+## What Levels 6–10 Would Add
+
+| Level | Name | What it unlocks |
+|-------|------|----------------|
+| 6 | Headless Mode | CLI scripts to run agents without interactive Claude, git hooks for auto-review |
+| 7 | Browser Automation | Puppeteer MCP for screenshots, PDF generation, web scraping |
+| 8 | Parallel Orchestration | Run multiple agents simultaneously (e.g., writer + typesetter in parallel) |
+| 9 | Background Agents | Cron jobs — nightly test builds, weekly reviews, memory refresh |
+| 10 | Self-Evolving | Autonomous improvement loop — agents that build and improve other agents |
+
+Run `/level-up` when ready to build Level 6.

package/package.json

@@ -0,0 +1,48 @@
+{
+  "name": "azrole",
+  "version": "3.0.0",
+  "description": "One command to build your entire AI coding environment with self-evolving intelligence. Works with Claude Code, Codex, OpenCode, Gemini CLI, and Cursor.",
+  "bin": {
+    "azrole": "./bin/cli.js",
+    "ai-dreamteam": "./bin/cli.js",
+    "claude-dreamteam": "./bin/cli.js"
+  },
+  "keywords": [
+    "azrole",
+    "claude",
+    "claude-code",
+    "codex",
+    "opencode",
+    "gemini",
+    "gemini-cli",
+    "cursor",
+    "ai",
+    "agents",
+    "developer-tools",
+    "automation",
+    "mcp",
+    "subagents",
+    "skills",
+    "orchestrator",
+    "ai-coding",
+    "vibe-coding",
+    "self-evolving",
+    "prompt-optimization",
+    "multi-agent"
+  ],
+  "author": "haytamAroui",
+  "license": "MIT",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/haytamAroui/claude-dreamteam.git"
+  },
+  "homepage": "https://github.com/haytamAroui/claude-dreamteam#readme",
+  "files": [
+    "bin/",
+    "templates/",
+    "docs/"
+  ],
+  "engines": {
+    "node": ">=16"
+  }
+}