NoizuPromptLingo: Command for Claude Code

Initialize Project

Bootstrap project documentation including CLAUDE.md, docs/PROJECT-ARCH.md, and docs/PROJECT-LAYOUT.md.

Phase 1: Initialize Environment

Initialize CLAUDE.md and load dependencies:

• Use the NPLLoad MCP tool to load NPL syntax elements: NPLLoad(expression="syntax fences directives")
• Use directory exploration tools (Glob, Bash tree) for directory structure exploration
• Use Read/Grep tools for file contents retrieval

CLAUDE.md receives:

• NPL Load Directive (environment variables, loading dependencies)
• NPL Scripts reference (dump-files, git-tree, npl-load, etc.)
• SQLite Quick Guide

---

Phase 2: Parallel Reconnaissance

Use @npl-project-coordinator to orchestrate parallel exploration.

Critical: Delegation-First Scanning

🎯 The coordinator MUST NOT perform deep codebase exploration directly.

Instead:

1. Quick Surface Scan — Coordinator performs only lightweight initial analysis:

   • ls top-level directories
   • Read package.json, Cargo.toml, or equivalent manifest
   • Check for existing README.md or docs
   • Identify project type and tech stack (~30 seconds max)

2. Immediate Delegation — Deploy scouts for all detailed exploration:

   • Do NOT recursively read files
   • Do NOT grep through entire codebase
   • Do NOT analyze individual source files
   • Let scouts handle domain-specific deep dives in parallel

3. Speed Through Parallelism — Launch all scouts simultaneously:

   • Each scout operates independently
   • Scouts return structured findings
   • Coordinator synthesizes results (doesn't re-scan)

Why: A coordinator scanning sequentially takes 10-20x longer than parallel scouts. The coordinator's job is orchestration, not exploration.

@npl-project-coordinator

## Task
Coordinate parallel reconnaissance of this codebase to generate PROJECT-ARCH.md and PROJECT-LAYOUT.md.

## Pre-Scout Analysis (30 seconds max)
- Identify project type from manifest files
- Note top-level structure
- Determine tech stack
- Then IMMEDIATELY deploy scouts

## Scout Deployment (Parallel)

Deploy 5-7 @npl-gopher-scout agents in parallel:

| Scout | Domain | Key Targets |
|-------|--------|-------------|
| Scout-Structure | File Organization | tree, files, naming conventions |
| Scout-Layers | Architecture | tiers, boundaries, data flow |
| Scout-Domain | Business Logic | bounded contexts, entities, aggregates |
| Scout-Patterns | Code Patterns | conventions, design patterns, testing |
| Scout-Services | Infrastructure | database, cache, queues, external APIs |
| Scout-API | Interfaces | endpoints, auth, contracts, versioning |
| Scout-DevOps | Operations | CI/CD, deployment, containers, IaC |

## Exploration Instructions

Each scout should:
1. Perform deep reconnaissance of their domain
2. Return structured findings with file:line references
3. Flag uncertainties and gaps
4. Estimate confidence per finding

## Execution Rules

🎯 **All scouts MUST be launched in a single parallel batch.**

- Use ONE Task tool call per scout, ALL in the same message
- Do NOT wait for one scout before launching the next
- Do NOT have the coordinator read files between scout launches
- Scouts operate independently - no inter-scout dependencies

## Synchronization
BARRIER: All scouts must complete before synthesis phase.