@vibe-agent-toolkit/vat-example-cat-agents 0.1.13 → 0.1.15-rc.1

package/dist/skills/vat-example-cat-agents/SKILL.md

@@ -52,8 +52,8 @@ Trigger this skill when:
 Use when user has a straightforward request that maps to one agent.
 
 **Example workflows:**
-- "What cat breed should I get?" → [breed-advisor](breed-advisor.md)
-- "Analyze this cat photo" → [photo-analyzer](photo-analyzer.md)
+- "What cat breed should I get?" → [breed-advisor](resources/breed-advisor.md)
+- "Analyze this cat photo" → [photo-analyzer](resources/photo-analyzer.md)
 - "Is this a valid haiku?" → haiku-validator (pure function)
 
 ### Pattern 2: Sequential Pipeline (Multi-Agent)
@@ -62,11 +62,11 @@ Use when output of one agent feeds into another.
 
 **Example workflows:**
 1. Photo → Characteristics → Name
-   - [photo-analyzer](photo-analyzer.md) → [name-generator](name-generator.md)
+   - [photo-analyzer](resources/photo-analyzer.md) → [name-generator](resources/name-generator.md)
 2. Photo → Characteristics → Haiku
-   - [photo-analyzer](photo-analyzer.md) → [haiku-generator](haiku-generator.md)
+   - [photo-analyzer](resources/photo-analyzer.md) → [haiku-generator](resources/haiku-generator.md)
 3. Description → Characteristics → Name → Validation
-   - [description-parser](description-parser.md) → [name-generator](name-generator.md) → name-validator
+   - [description-parser](resources/description-parser.md) → [name-generator](resources/name-generator.md) → name-validator
 
 ### Pattern 3: Generate-Validate Loop (Iterative)
 
@@ -75,10 +75,10 @@ Use when generator produces content that needs validation with retry logic.
 **Example workflows:**
 1. Name generation with validation
    - Generate → Validate → If invalid, retry with feedback
-   - Uses: [name-generator](name-generator.md) + name-validator
+   - Uses: [name-generator](resources/name-generator.md) + name-validator
 2. Haiku generation with validation
    - Generate → Validate → If invalid, retry
-   - Uses: [haiku-generator](haiku-generator.md) + haiku-validator
+   - Uses: [haiku-generator](resources/haiku-generator.md) + haiku-validator
 
 **Orchestration tip:** Generator agents have NO knowledge of validation rules. This is intentional - forces iteration and tests feedback loops.
 
@@ -88,7 +88,7 @@ Use when decision requires human judgment.
 
 **Example workflows:**
 1. Breed application approval
-   - Gather info → Generate application → [human-approval](human-approval.md) → Process result
+   - Gather info → Generate application → [human-approval](resources/human-approval.md) → Process result
 2. Name approval before finalization
    - Generate names → Present options → Human selects → Finalize
 
@@ -98,10 +98,10 @@ Use when decision requires human judgment.
 
 **When:** User wants help finding the right cat breed.
 
-**Agent:** [breed-advisor](breed-advisor.md)
+**Agent:** [breed-advisor](resources/breed-advisor.md)
 
 **Orchestration strategy:**
-1. **Phase 1 - Gathering** (see [Conversation Strategy](breed-advisor.md#conversation-strategy))
+1. **Phase 1 - Gathering** (see [Conversation Strategy](resources/breed-advisor.md#conversation-strategy))
    - Collect ≥4 factors including music preference
    - ONE question at a time (don't bombard)
    - Extract factors after each turn
@@ -117,15 +117,15 @@ Use when decision requires human judgment.
    - Conclude gracefully
    - Provide next steps or exit instruction
 
-**Critical factor:** Music preference is the PRIMARY compatibility factor. Ask early, use as conversation anchor. See [Music Preference Insight](breed-advisor.md#music-preference-insight) for mappings.
+**Critical factor:** Music preference is the PRIMARY compatibility factor. Ask early, use as conversation anchor. See [Music Preference Insight](resources/breed-advisor.md#music-preference-insight) for mappings.
 
-**Reference implementation:** See [cat-breed-selection.md](cat-breed-selection.md) for detailed breed selection orchestration.
+**Reference implementation:** See [cat-breed-selection.md](resources/cat-breed-selection.md) for detailed breed selection orchestration.
 
 ### Workflow B: Photo Analysis Pipeline
 
 **When:** User provides a cat photo and wants analysis, name, or haiku.
 
-**Agents:** [photo-analyzer](photo-analyzer.md) → [name-generator](name-generator.md) OR [haiku-generator](haiku-generator.md)
+**Agents:** [photo-analyzer](resources/photo-analyzer.md) → [name-generator](resources/name-generator.md) OR [haiku-generator](resources/haiku-generator.md)
 
 **Orchestration strategy:**
 
@@ -154,7 +154,7 @@ Optional Step 3: Validate
 
 **When:** User describes a cat in text (no photo).
 
-**Agents:** [description-parser](description-parser.md) → [name-generator](name-generator.md) OR [haiku-generator](haiku-generator.md)
+**Agents:** [description-parser](resources/description-parser.md) → [name-generator](resources/name-generator.md) OR [haiku-generator](resources/haiku-generator.md)
 
 **Orchestration strategy:**
 
@@ -210,14 +210,14 @@ throw new Error('Could not generate valid content after 3 attempts');
 - ~60-70% initial rejection rate forces iteration
 
 **Agents using this pattern:**
-- [name-generator](name-generator.md) + name-validator
-- [haiku-generator](haiku-generator.md) + haiku-validator
+- [name-generator](resources/name-generator.md) + name-validator
+- [haiku-generator](resources/haiku-generator.md) + haiku-validator
 
 ### Workflow E: HITL Approval Gate
 
 **When:** Decision requires human judgment (compliance, taste, ethics).
 
-**Agent:** [human-approval](human-approval.md)
+**Agent:** [human-approval](resources/human-approval.md)
 
 **Orchestration strategy:**
 
@@ -504,30 +504,30 @@ Good: "What's your favorite type of music?" → [response] → "Great! Tell me a
 - haiku-validator (no agent resource - pure TypeScript logic)
 
 ### One-Shot LLM Analyzers
-- [Photo Analyzer](photo-analyzer.md) - Vision analysis
-- [Description Parser](description-parser.md) - Text parsing
-- [Name Generator](name-generator.md) - Creative naming
-- [Haiku Generator](haiku-generator.md) - Haiku composition
+- [Photo Analyzer](resources/photo-analyzer.md) - Vision analysis
+- [Description Parser](resources/description-parser.md) - Text parsing
+- [Name Generator](resources/name-generator.md) - Creative naming
+- [Haiku Generator](resources/haiku-generator.md) - Haiku composition
 
 ### Conversational Assistant
-- [Breed Advisor](breed-advisor.md) - Multi-turn breed selection
-  - [Welcome Message](breed-advisor.md#welcome-message)
-  - [Music Preference Insight](breed-advisor.md#music-preference-insight)
-  - [Factor Definitions](breed-advisor.md#factor-definitions)
-  - [Conversation Strategy](breed-advisor.md#conversation-strategy)
-  - [Factor Extraction Prompt](breed-advisor.md#factor-extraction-prompt)
-  - [Transition Message](breed-advisor.md#transition-message)
-  - [Recommendation Presentation](breed-advisor.md#recommendation-presentation-prompt)
-  - [Selection Extraction](breed-advisor.md#selection-extraction-prompt)
-  - [Conclusion Prompt](breed-advisor.md#conclusion-prompt)
+- [Breed Advisor](resources/breed-advisor.md) - Multi-turn breed selection
+  - [Welcome Message](resources/breed-advisor.md#welcome-message)
+  - [Music Preference Insight](resources/breed-advisor.md#music-preference-insight)
+  - [Factor Definitions](resources/breed-advisor.md#factor-definitions)
+  - [Conversation Strategy](resources/breed-advisor.md#conversation-strategy)
+  - [Factor Extraction Prompt](resources/breed-advisor.md#factor-extraction-prompt)
+  - [Transition Message](resources/breed-advisor.md#transition-message)
+  - [Recommendation Presentation](resources/breed-advisor.md#recommendation-presentation-prompt)
+  - [Selection Extraction](resources/breed-advisor.md#selection-extraction-prompt)
+  - [Conclusion Prompt](resources/breed-advisor.md#conclusion-prompt)
 
 ### External Event Integrator
-- [Human Approval](human-approval.md) - HITL approval gate
+- [Human Approval](resources/human-approval.md) - HITL approval gate
 
 ### Related Documentation
-- [Cat Breed Selection Skill](cat-breed-selection.md) - Detailed breed advisor orchestration
+- [Cat Breed Selection Skill](resources/cat-breed-selection.md) - Detailed breed advisor orchestration
 - Package README - Human-facing documentation
-- Package Structure - Technical organization
+- [Package Structure](resources/structure.md) - Technical organization
 
 ## Success Criteria
 

package/dist/skills/vat-example-cat-agents/{human-approval.md → resources/human-approval.md}

@@ -333,5 +333,5 @@ it('should timeout after specified duration', async () => {
 
 ## Related Documentation
 
-- **[SKILL.md](SKILL.md#workflow-e-hitl-approval-gate)** - Orchestration patterns
-- **[Agent Runtime](../../../agent-runtime/README.md)** - executeExternalEvent helper
+- **[SKILL.md](../SKILL.md#workflow-e-hitl-approval-gate)** - Orchestration patterns
+- **Agent Runtime** - executeExternalEvent helper

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vibe-agent-toolkit/vat-example-cat-agents",
-  "version": "0.1.13",
+  "version": "0.1.15-rc.1",
   "description": "Example agents: 8 quirky cat agents demonstrating VAT patterns",
   "type": "module",
   "exports": {
@@ -37,24 +37,21 @@
     "postinstall": "vat skills install --npm-postinstall || exit 0"
   },
   "dependencies": {
-    "@vibe-agent-toolkit/agent-runtime": "0.1.13",
-    "@vibe-agent-toolkit/agent-schema": "0.1.13",
-    "@vibe-agent-toolkit/transports": "0.1.13",
-    "@vibe-agent-toolkit/utils": "0.1.13",
+    "@vibe-agent-toolkit/agent-runtime": "0.1.15-rc.1",
+    "@vibe-agent-toolkit/agent-schema": "0.1.15-rc.1",
+    "@vibe-agent-toolkit/transports": "0.1.15-rc.1",
     "syllable": "^5.0.1",
     "zod": "^3.24.1"
   },
   "devDependencies": {
     "@ai-sdk/openai": "^3.0.18",
-    "@anthropic-ai/claude-agent-sdk": "^0.1.5",
     "@anthropic-ai/sdk": "^0.36.0",
     "@langchain/openai": "^0.3.16",
     "@types/node": "^22.10.5",
-    "@vibe-agent-toolkit/resource-compiler": "0.1.13",
-    "@vibe-agent-toolkit/runtime-claude-agent-sdk": "0.1.13",
-    "@vibe-agent-toolkit/runtime-langchain": "0.1.13",
-    "@vibe-agent-toolkit/runtime-openai": "0.1.13",
-    "@vibe-agent-toolkit/runtime-vercel-ai-sdk": "0.1.13",
+    "@vibe-agent-toolkit/resource-compiler": "0.1.15-rc.1",
+    "@vibe-agent-toolkit/runtime-claude-agent-sdk": "0.1.15-rc.1",
+    "@vibe-agent-toolkit/runtime-langchain": "0.1.15-rc.1",
+    "@vibe-agent-toolkit/runtime-openai": "0.1.15-rc.1",
     "@vitest/coverage-v8": "2.1.8",
     "ai": "^6.0.48",
     "openai": "^4.77.3",

package/dist/skills/vat-example-cat-agents/CLAUDE.md

@@ -1,424 +0,0 @@
-# Claude Code Navigation Guide
-
-This file provides technical navigation details for AI assistants working on `vat-example-cat-agents`.
-
-## Quick Reference
-
-**Package Purpose:** Reference implementation demonstrating VAT agent patterns across 9 archetypes using cat breeding domain.
-
-**Current Status:** 4 of 9 archetypes implemented (8 agents total)
-
-**Key Principle:** Agents are plain TypeScript functions. No framework dependencies in agent code.
-
-## Directory Navigation
-
-```
-packages/vat-example-cat-agents/
-├── src/
-│   ├── types/schemas.ts               # Shared Zod schemas (CatCharacteristics, Haiku, etc.)
-│   │
-│   ├── pure-function-tool/            # Archetype 1 (2 agents)
-│   │   ├── haiku-validator.ts         # Validates 5-7-5 syllable structure + kigo/kireji
-│   │   └── name-validator.ts          # Quirky characteristic-based validation
-│   │
-│   ├── one-shot-llm-analyzer/         # Archetype 2 (4 agents)
-│   │   ├── photo-analyzer.ts          # Vision LLM → CatCharacteristics
-│   │   ├── description-parser.ts      # Text → CatCharacteristics (same schema as photo)
-│   │   ├── name-generator.ts          # CatCharacteristics → NameSuggestion
-│   │   └── haiku-generator.ts         # CatCharacteristics → Haiku
-│   │
-│   ├── conversational-assistant/      # Archetype 3 (1 agent)
-│   │   ├── breed-advisor.ts           # Multi-turn breed selection advisor
-│   │   └── breed-knowledge.ts         # Breed database and matching algorithm
-│   │
-│   ├── external-event-integrator/     # Archetype 9 (1 agent)
-│   │   └── human-approval.ts          # HITL approval gate (mockable)
-│   │
-│   ├── utils/
-│   │   └── color-extraction.ts        # Helper: Extract colors from descriptions
-│   │
-│   └── index.ts                       # Public exports
-│
-├── test/
-│   ├── pure-function-tool/            # Unit tests for validators
-│   ├── one-shot-llm-analyzer/         # Tests for LLM analyzers
-│   ├── external-event-integrator/     # Tests for HITL
-│   ├── fixtures/
-│   │   └── photos/                    # Test images with EXIF metadata
-│   │       ├── cats/                  # 4 cat photos (~40KB each)
-│   │       ├── not-cats/              # 2 negative cases (bear, robot)
-│   │       └── cat-like/              # Future: ambiguous cases
-│   └── test-helpers.ts                # Shared test utilities
-│
-├── examples/
-│   ├── photo-analysis-demo.ts         # Demos photo analyzer with test fixtures
-│   ├── runtime-adapter-demo.ts        # Shared runtime adapter demo
-│   ├── llm-agent-demo.ts              # LLM agent usage patterns
-│   └── demo-helpers.ts                # Terminal color output utilities
-│
-├── README.md                          # Human-facing documentation
-├── CLAUDE.md                          # This file (AI navigation)
-├── docs/structure.md                       # Package organization details
-└── package.json
-```
-
-## Agent Organization by Archetype
-
-### Implemented Archetypes
-
-**Archetype 1: Pure Function Tool**
-- Location: `src/pure-function-tool/`
-- Characteristics: Stateless, deterministic, no external dependencies
-- Agents:
-  - `haiku-validator.ts` - Syllable counting + kigo/kireji detection
-  - `name-validator.ts` - Characteristic-based quirky rules (60-70% rejection rate)
-
-**Archetype 2: One-Shot LLM Analyzer**
-- Location: `src/one-shot-llm-analyzer/`
-- Characteristics: Single LLM call, stateless, classification/extraction
-- Agents:
-  - `photo-analyzer.ts` - Vision LLM (mockable via EXIF)
-  - `description-parser.ts` - Text parsing (multi-modal convergence with photo)
-  - `name-generator.ts` - Creative name generation
-  - `haiku-generator.ts` - Haiku creation
-
-**Archetype 3: Conversational Assistant**
-- Location: `src/conversational-assistant/`
-- Characteristics: Multi-turn conversation, session state, context accumulation
-- Agents:
-  - `breed-advisor.ts` - Breed selection advisor with music-based matching
-  - `breed-knowledge.ts` - Breed database with 12 breeds and matching algorithm
-
-**Archetype 9: External Event Integrator**
-- Location: `src/external-event-integrator/`
-- Characteristics: Emits events, blocks waiting, timeout handling
-- Agents:
-  - `human-approval.ts` - HITL approval (mockable)
-
-### Missing Archetypes (Planned)
-
-**Archetype 4: Agentic Researcher (ReAct)** ⏸️
-- Needs: Tool calling, iterative reasoning
-- Target: Breed history researcher with web search
-
-**Archetype 5: Function Workflow Orchestrator** ⏸️
-- Needs: Deterministic multi-agent coordination
-- Target: Breeding approval pipeline
-- NOTE: Previous implementation removed during AI SDK v6 migration
-
-**Archetype 6: LLM Intelligent Coordinator** ⏸️
-- Needs: LLM routing decisions in fixed workflow
-- Target: Smart submission router
-
-**Archetype 7: Function Event Consumer** ⏸️
-- Needs: Event-triggered execution
-- Target: Pedigree file processor
-
-**Archetype 8: LLM Event Handler** ⏸️
-- Needs: Event-triggered with LLM processing
-- Target: Intelligent triage handler
-
-## Code Patterns
-
-### Agent File Structure
-
-Each agent file exports both the function and metadata:
-
-```typescript
-// src/<archetype>/<agent-name>.ts
-
-import { z } from 'zod';
-import { InputSchema, OutputSchema } from '../types/schemas.js';
-
-/**
- * Agent function implementation
- */
-export async function agentFunction(input: z.infer<typeof InputSchema>): Promise<z.infer<typeof OutputSchema>> {
-  // Implementation
-}
-
-/**
- * Agent metadata (for future VAT integration)
- */
-export const AgentMetadata = {
-  id: 'archetype/agent-name',
-  name: 'Human Readable Name',
-  description: 'What this agent does',
-  archetype: 'archetype-name',
-  inputSchema: InputSchema,
-  outputSchema: OutputSchema,
-};
-```
-
-### Multi-Modal Convergence Pattern
-
-Photo Analyzer and Description Parser both produce `CatCharacteristics`:
-
-```typescript
-// Both functions return the same schema
-const fromImage = await analyzePhoto(imagePath);   // Vision LLM
-const fromText = await parseDescription(text);     // Text parsing LLM
-
-// Downstream agents accept either
-const name = await generateCatName(fromImage);     // Works
-const name = await generateCatName(fromText);      // Also works
-```
-
-### Mockable Pattern (Testing)
-
-Agents that call expensive APIs support mock mode:
-
-```typescript
-// photo-analyzer.ts
-export async function analyzePhoto(
-  imagePath: string,
-  options: { mockable?: boolean } = {}
-): Promise<CatCharacteristics> {
-  const { mockable = true } = options;
-
-  if (mockable) {
-    // Read from EXIF metadata (fast, free, deterministic)
-    return extractFromExif(imagePath);
-  }
-
-  // Call real vision API (slow, costs money)
-  return callVisionAPI(imagePath);
-}
-```
-
-### Feedback Loop Pattern
-
-Generator + Validator creates iteration pattern:
-
-```typescript
-// Generator has NO knowledge of validation rules
-const suggestion = await generateCatName(characteristics);
-
-// Validator rejects ~60-70% (quirky rules)
-const validation = validateCatName(suggestion.name, characteristics);
-
-if (validation.status === 'invalid') {
-  // Retry with different approach
-  // Tests multi-turn orchestration
-}
-```
-
-## Shared Schemas
-
-All agents use schemas from `src/types/schemas.ts`:
-
-### Core Schemas
-
-- **CatCharacteristics** - Physical + behavioral traits
-- **Haiku** - Three-line poem structure
-- **NameSuggestion** - Name + reasoning + alternatives
-- **NameValidationResult** - Approval/rejection with reason
-- **HaikuValidationResult** - Syllable counts + errors + kigo/kireji
-
-### Schema Relationships
-
-```
-Photo/Text Input
-      ↓
-CatCharacteristics ────┬─→ Name Generator → NameSuggestion → Name Validator
-                       │
-                       └─→ Haiku Generator → Haiku → Haiku Validator
-```
-
-## Test Fixtures Strategy
-
-### Images with EXIF Metadata
-
-**Problem:** Vision APIs are slow and expensive for testing.
-
-**Solution:** Embed test expectations in EXIF metadata.
-
-**Process:**
-1. Download cat photos from Unsplash (~1-9MB each)
-2. Run `process-test-images.ts` utility:
-   - Resizes to 512px wide
-   - Compresses to ~50KB
-   - Extracts metadata from filename patterns
-   - Writes JSON to EXIF Description field
-3. Commit processed images to git (~40KB each)
-
-**Mock Mode:**
-- Reads EXIF metadata instead of calling vision API
-- Fast, free, deterministic
-- Perfect for CI/CD
-
-**Real Mode:**
-- Set `mockable: false`
-- Calls actual vision API
-- Slow, costs money, non-deterministic
-- Use sparingly for integration tests
-
-### Filename Patterns
-
-The `process-test-images.ts` utility auto-detects from filenames:
-
-- **Colors:** `orange`, `black`, `white`, `gray`, `calico`
-- **Patterns:** `tabby`, `solid`, `patched`, `striped`
-- **Sizes:** `tiny`, `small`, `large` (default: `medium`)
-- **Breeds:** `persian`, `maine-coon`, `siamese`, `domestic-shorthair`
-- **Personality:** `playful`, `lazy`, `grumpy`, `affectionate`, `curious`, `regal`
-- **Quirks:** `three-leg`, `cross-eye`, `scar`
-
-**Example:** `orange-tabby-playful.jpg` → Orange tabby with playful personality
-
-## Running Tests
-
-```bash
-# All tests
-bun run test
-
-# Watch mode
-bun run test:watch
-
-# Coverage
-bun run test:coverage
-
-# Specific archetype
-bun run test src/pure-function-tool
-```
-
-## Running Demos
-
-```bash
-# Photo analysis demo (uses test fixtures)
-bun run demo:photos
-
-# Runtime adapter demos (in other packages)
-cd packages/runtime-vercel-ai-sdk && bun run demo
-cd packages/runtime-langchain && bun run demo
-```
-
-## Adding New Agents
-
-### Step 1: Choose Archetype
-
-Determine which of the 9 archetypes fits your agent.
-
-### Step 2: Create Agent File
-
-```bash
-# Create in appropriate archetype directory
-touch src/<archetype>/<agent-name>.ts
-```
-
-### Step 3: Implement Agent
-
-Follow the agent file structure pattern (see Code Patterns above).
-
-### Step 4: Add Tests
-
-```bash
-# Create test file
-touch test/<archetype>/<agent-name>.test.ts
-```
-
-### Step 5: Export from Index
-
-```typescript
-// src/index.ts
-export * from './<archetype>/<agent-name>.js';
-```
-
-### Step 6: Update README
-
-Add agent documentation to README.md under appropriate archetype section.
-
-## Common Tasks
-
-### Add Test Fixture Images
-
-```bash
-# 1. Download images to temp directory
-# 2. Process images
-cd packages/dev-tools
-bun run process-images ~/Downloads/cat-photos ../../vat-example-cat-agents/test/fixtures/photos/cats
-
-# 3. Review output (should be <100KB per image)
-# 4. Commit processed images
-```
-
-### Add New Schema
-
-```typescript
-// src/types/schemas.ts
-export const NewSchema = z.object({
-  // Define schema
-}).describe('Schema description');
-
-export type NewType = z.infer<typeof NewSchema>;
-```
-
-### Mock LLM Calls in Tests
-
-```typescript
-// test/<archetype>/<agent>.test.ts
-import { vi } from 'vitest';
-
-// Mock the LLM call
-vi.mock('../src/one-shot-llm-analyzer/photo-analyzer', () => ({
-  analyzePhoto: vi.fn().mockResolvedValue({
-    physical: { furColor: 'Orange', size: 'medium' },
-    behavioral: { personality: ['Playful'] },
-    description: 'Test description',
-  }),
-}));
-```
-
-## Architecture Decisions
-
-### Why Plain TypeScript Functions?
-
-- **Framework agnostic** - No vendor lock-in
-- **Easy to test** - Pure functions, no magic
-- **Portable** - Runtime adapters translate to any framework
-- **Simple** - No framework abstractions to learn
-
-### Why Mock Mode?
-
-- **Speed** - Instant vs seconds/minutes for real API
-- **Cost** - Free vs $0.01-0.10 per image
-- **Deterministic** - Same input = same output
-- **CI/CD friendly** - No API keys, no rate limits
-
-### Why Cat Domain?
-
-- **Universal** - Everyone understands cats
-- **Complex enough** - Genetics, breeding, compliance
-- **Fun** - Humor makes exploration enjoyable
-- **Realistic** - Tests real-world patterns (HITL, workflows, etc.)
-
-### Why Quirky Validation Rules?
-
-- **Forces iteration** - 60-70% rejection rate tests feedback loops
-- **Tests orchestration** - Generator doesn't know rules
-- **Prevents cheating** - LLM can't rely on training (rules are arbitrary)
-- **Demonstrates patterns** - Real-world validation is often quirky
-
-## Related Documentation
-
-- **README.md** - Human-facing package documentation
-- **docs/structure.md** - Detailed package organization
-- **[Project CLAUDE.md](../../CLAUDE.md)** - Root-level project guidance
-
-## Contributing Workflow
-
-1. Review the 9 agent archetypes (see Missing Archetypes section above)
-2. Choose an unimplemented archetype
-3. Create agent file(s) in appropriate directory
-4. Add tests with good coverage
-5. Update README.md with usage examples
-6. Export from src/index.ts
-7. Run `bun run validate` to ensure everything passes
-8. Commit with conventional commits format
-
-## Questions?
-
-For high-level understanding: Read README.md
-For technical navigation: Read this file (CLAUDE.md)
-For package structure details: Read docs/structure.md
-For archetype theory: See the Missing Archetypes section above

package/dist/skills/vat-example-cat-agents/README.md

@@ -1,727 +0,0 @@
-# @vibe-agent-toolkit/vat-example-cat-agents
-
-**Example agents demonstrating VAT patterns across 9 agent archetypes.**
-
-## Purpose
-
-This package serves as a **reference implementation** for the Vibe Agent Toolkit (VAT). It demonstrates how to build portable agents that work across multiple frameworks (Vercel AI SDK, LangChain, Claude Agent SDK, n8n) using a whimsical but realistic cat breeding domain.
-
-**Why cats?** The domain provides realistic complexity (genetics, compliance, multi-step workflows) while remaining approachable and fun. Think of it as the "TodoMVC" of agentic AI frameworks.
-
-## Current Status
-
-**Implementation Progress:** 4 of 9 archetypes, 8 agents
-
-| Archetype | Status | Agents |
-|-----------|--------|--------|
-| 1. Pure Function Tool | ✅ Complete | 2 agents |
-| 2. One-Shot LLM Analyzer | ✅ Complete | 4 agents |
-| 3. Conversational Assistant | ✅ Complete | 1 agent |
-| 4. Agentic Researcher (ReAct) | ⏸️ Planned | 0 agents |
-| 5. Function Workflow Orchestrator | ⏸️ Planned | 0 agents |
-| 6. LLM Intelligent Coordinator | ⏸️ Planned | 0 agents |
-| 7. Function Event Consumer | ⏸️ Planned | 0 agents |
-| 8. LLM Event Handler | ⏸️ Planned | 0 agents |
-| 9. External Event Integrator | ✅ Complete | 1 agent |
-
-**Note:** This is a living example package. Archetypes are implemented as we validate the VAT framework design.
-
-## Installation
-
-### As NPM Package
-
-```bash
-npm install @vibe-agent-toolkit/vat-example-cat-agents
-```
-
-### Installing the Skill
-
-This package includes a distributable Claude Code skill called `vat-cat-agents` that provides orchestration guidance for using the 8 example cat agents.
-
-**Install from npm:**
-```bash
-vat skills install npm:@vibe-agent-toolkit/vat-example-cat-agents
-```
-
-**Install from local directory (development):**
-```bash
-vat skills install ./packages/vat-example-cat-agents
-```
-
-**Verify installation:**
-```bash
-vat skills list --installed
-```
-
-The `vat-cat-agents` skill will be installed to `~/.claude/plugins/vat-cat-agents/` and will appear in Claude Code after restarting or running `/reload-skills`.
-
-**What the skill includes:**
-- SKILL.md with agent orchestration patterns
-- Documentation on multi-agent workflows
-- Examples of feedback loops (generator → validator)
-- Multi-modal input patterns (photo vs text)
-- Links to all 8 agent implementations
-
-## Result Envelopes
-
-All VAT agents return standardized result envelopes following Railway-Oriented Programming (ROP) principles. This provides consistent error handling, type-safe result processing, and clear orchestration patterns.
-
-### OneShotAgentOutput (Pure Functions & One-Shot LLM)
-
-Pure function tools and one-shot LLM analyzers return `OneShotAgentOutput<TData, TError>`:
-
-```typescript
-interface OneShotAgentOutput<TData, TError> {
-  result: AgentResult<TData, TError>;
-}
-
-type AgentResult<TData, TError> =
-  | { status: 'success'; data: TData }
-  | { status: 'error'; error: TError };
-```
-
-**Example:**
-```typescript
-const output = await haikuValidator.execute({ text, syllables, kigo, kireji });
-
-if (output.result.status === 'success') {
-  console.log('Valid:', output.result.data.valid);
-} else {
-  console.error('Invalid:', output.result.error);
-}
-```
-
-### ConversationalAgentOutput (Multi-Turn Agents)
-
-Conversational assistants return `ConversationalAgentOutput<TData, TError, TState>`:
-
-```typescript
-interface ConversationalAgentOutput<TData, TError, TState> {
-  reply: string;                // Natural language response
-  sessionState: TState;         // Updated session state
-  result: StatefulAgentResult<TData, TError, TMetadata>;
-}
-
-type StatefulAgentResult<TData, TError, TMetadata> =
-  | { status: 'in-progress'; metadata?: TMetadata }
-  | { status: 'success'; data: TData }
-  | { status: 'error'; error: TError };
-```
-
-**Example:**
-```typescript
-const output = await breedAdvisor.execute({
-  message: 'I love classical music',
-  sessionState: { profile: { conversationPhase: 'gathering' } },
-});
-
-console.log('Agent says:', output.reply);           // For user
-const nextState = output.sessionState;              // Carry to next turn
-
-if (output.result.status === 'in-progress') {
-  console.log('Gathering info:', output.result.metadata);
-} else if (output.result.status === 'success') {
-  console.log('Recommendation:', output.result.data);
-}
-```
-
-See [Orchestration Guide](../../docs/orchestration.md) for detailed patterns and examples.
-
-## Agent Catalog
-
-### Archetype 1: Pure Function Tool
-
-**Characteristics:** Stateless, synchronous, deterministic, no external dependencies
-
-**Use Cases:** Validation, calculation, formatting, rule-based logic
-
-#### Haiku Validator
-
-Validates haiku structure according to traditional Japanese poetry rules.
-
-```typescript
-import { validateHaiku, type Haiku } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-const haiku: Haiku = {
-  line1: 'Autumn moon rises',
-  line2: 'Silver light on quiet waves',
-  line3: 'The cat sits and waits',
-};
-
-const result = validateHaiku(haiku);
-console.log(result);
-// {
-//   valid: true,
-//   syllables: { line1: 5, line2: 7, line3: 5 },
-//   errors: [],
-//   hasKigo: true,    // Seasonal reference detected
-//   hasKireji: false  // No cutting word
-// }
-```
-
-**Validation Rules:**
-- 5-7-5 syllable structure (strict)
-- Kigo detection (seasonal words: spring, autumn, winter, summer, etc.)
-- Kireji detection (cutting words: や, かな, けり, etc.)
-
-**Not Cat-Specific:** This is a general-purpose haiku validator, reusable for any haiku validation.
-
-#### Name Validator
-
-Validates cat names against whimsical characteristic-based rules.
-
-```typescript
-import { validateCatName, type CatCharacteristics } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-const cat: CatCharacteristics = {
-  physical: {
-    furColor: 'Orange',
-    furPattern: 'Tabby',
-    eyeColor: 'Green',
-    size: 'large',
-  },
-  behavioral: {
-    personality: ['Regal', 'Demanding'],
-  },
-  description: 'A large orange tabby who rules the household',
-};
-
-const result = validateCatName('Duke Marmalade III', cat);
-console.log(result);
-// {
-//   status: 'valid',
-//   reason: 'Proper masculine nobility with food-related theme!'
-// }
-```
-
-**Quirky Validation Rules:**
-- Three-legged cats must have three-syllable names
-- Black cats cannot have names containing the letter 'e'
-- Orange cats must have food-related names (Marmalade, Pumpkin, etc.)
-- High-energy cats need short names (≤5 letters)
-- Fluffy cats need repeated consonants (Mittens, Fluffy, etc.)
-
-**Purpose:** Tests feedback loops (60-70% rejection rate forces iteration patterns)
-
----
-
-### Archetype 2: One-Shot LLM Analyzer
-
-**Characteristics:** Single LLM call, no iteration, stateless, classification/extraction/generation
-
-**Use Cases:** Image analysis, text parsing, classification, structured extraction, generation
-
-#### Photo Analyzer
-
-Extracts structured cat characteristics from images using vision LLM.
-
-```typescript
-import { analyzePhoto } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-const characteristics = await analyzePhoto('/path/to/cat-photo.jpg');
-console.log(characteristics);
-// {
-//   physical: {
-//     furColor: 'Orange',
-//     furPattern: 'Tabby',
-//     eyeColor: 'Green',
-//     breed: 'Domestic Shorthair',
-//     size: 'medium'
-//   },
-//   behavioral: {
-//     personality: ['Playful', 'Curious'],
-//     quirks: []
-//   },
-//   description: 'An orange tabby cat with green eyes...'
-// }
-```
-
-**Mock Mode:** Default behavior extracts from EXIF metadata + filename patterns (fast, free, deterministic). Set `mockable: false` to use real vision API.
-
-**Multi-Modal Input:** Produces same `CatCharacteristics` schema as Description Parser (interchangeable inputs).
-
-#### Description Parser
-
-Parses unstructured text descriptions into structured cat characteristics.
-
-```typescript
-import { parseDescription } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-const text = "Fluffy is a large, playful orange tabby with green eyes. She's very curious and loves to explore.";
-const characteristics = await parseDescription(text);
-// Returns same CatCharacteristics schema as Photo Analyzer
-```
-
-**Multi-Modal Convergence:** Text input → same schema as image input. Enables pipelines that accept either photos or descriptions.
-
-#### Name Generator
-
-Generates creative cat names based on characteristics.
-
-```typescript
-import { generateCatName } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-const name = await generateCatName(characteristics);
-console.log(name);
-// {
-//   name: 'Duke Marmalade III',
-//   reasoning: 'Orange color suggests food theme, regal personality demands nobility',
-//   alternatives: ['Sir Butterscotch', 'Lord Pumpkin']
-// }
-```
-
-**No Knowledge of Rules:** Generator does NOT know the validation rules. This is intentional - tests feedback loop patterns where generator → validator → retry.
-
-#### Haiku Generator
-
-Creates cat-themed haikus from characteristics.
-
-```typescript
-import { generateHaiku } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-const haiku = await generateHaiku(characteristics);
-console.log(haiku);
-// {
-//   line1: 'Orange fur gleaming',
-//   line2: 'Playful paws dance in sunshine',
-//   line3: 'Green eyes watch and wait'
-// }
-```
-
-**Feedback Loop:** Generated haikus can be validated with Haiku Validator. Tests iteration patterns.
-
----
-
-### Archetype 3: Conversational Assistant
-
-**Characteristics:** Multi-turn conversation, session state management, context accumulation, flexible dialogue
-
-**Use Cases:** Advisory chatbots, guided workflows, interactive Q&A, personalized recommendations
-
-#### Breed Selection Advisor
-
-Helps users find their perfect cat breed through flexible, natural conversation. Tracks selection factors across multiple turns and provides personalized recommendations based on a whimsical music-breed compatibility theory.
-
-```typescript
-import { breedAdvisorAgent } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-// Turn 1: Initial greeting
-let result = await breedAdvisorAgent({
-  message: "Hi! I'm looking for a cat breed that fits my lifestyle.",
-  sessionState: undefined, // No prior state
-});
-
-console.log(result.reply);
-// "Hello! I'd love to help you find the perfect breed! To get started, tell me..."
-console.log(result.updatedProfile.conversationPhase); // "gathering"
-
-// Turn 2: Provide information
-result = await breedAdvisorAgent({
-  message: "I live in a small apartment and work from home. I listen to jazz music while coding.",
-  sessionState: {
-    profile: result.updatedProfile, // Pass updated profile from previous turn
-  },
-});
-
-console.log(result.reply);
-// "Excellent! Jazz lovers tend to appreciate cats with..."
-console.log(result.updatedProfile);
-// {
-//   livingSpace: 'apartment',
-//   musicPreference: 'jazz',
-//   conversationPhase: 'gathering'
-// }
-
-// Turn 3: More information
-result = await breedAdvisorAgent({
-  message: "I prefer low-maintenance grooming and I'm moderately active.",
-  sessionState: {
-    profile: result.updatedProfile,
-  },
-});
-
-console.log(result.updatedProfile.conversationPhase); // "ready-to-recommend"
-console.log(result.recommendations);
-// [
-//   {
-//     breed: 'Siamese',
-//     matchScore: 85,
-//     reasoning: 'music preference (jazz) aligns perfectly; activity level (playful-moderate) matches well; suitable for apartment living'
-//   },
-//   {
-//     breed: 'Bengal',
-//     matchScore: 70,
-//     reasoning: 'music preference (jazz) aligns perfectly; grooming needs (minimal) match tolerance'
-//   },
-//   // ... more recommendations
-// ]
-
-// Turn 4: Follow-up question
-result = await breedAdvisorAgent({
-  message: "Tell me more about the Siamese - are they vocal?",
-  sessionState: {
-    profile: result.updatedProfile,
-  },
-});
-
-console.log(result.updatedProfile.conversationPhase); // "refining"
-// Agent provides detailed information about Siamese cats
-```
-
-**Selection Factors Tracked:**
-
-1. **Living Space** - apartment, small-house, large-house, farm
-2. **Activity Level** - couch-companion, playful-moderate, active-explorer, high-energy-athlete
-3. **Grooming Tolerance** - minimal, weekly, daily
-4. **Family Composition** - single, couple, young-kids, older-kids, multi-pet
-5. **Allergies** - boolean (requires hypoallergenic breeds)
-6. **Music Preference** - classical, jazz, rock, metal, pop, country, electronic, none (CRITICAL!)
-
-**The Music Factor (CRITICAL!):**
-
-This agent is built on a whimsical theory: **music preference is the MOST IMPORTANT factor** in cat breed selection!
-
-Each music genre aligns with specific breed temperaments through "vibrational frequency compatibility":
-
-- **Classical** → Calm, regal breeds (Persian, Ragdoll) - harmonic resonance
-- **Jazz** → Intelligent, unpredictable breeds (Siamese, Bengal) - improvisational energy
-- **Rock/Metal** → High-energy, bold breeds (Maine Coon, Abyssinian) - intensity matching
-- **Pop** → Social, adaptable breeds (Domestic Shorthair) - mainstream compatibility
-- **Electronic** → Modern, quirky breeds (Sphynx, Devon Rex) - synthetic-natural balance
-- **Country** → Traditional, loyal breeds (American Shorthair) - heartland values
-- **None/Silence** → Independent, mysterious breeds (Russian Blue) - zen alignment
-
-Music preference receives **30 points (30% weight)** in the matching algorithm - more than any other factor!
-
-**Conversation Phases:**
-
-The agent automatically transitions through three phases based on information gathered:
-
-1. **Gathering** (<4 factors): Asks questions to collect missing factors, prioritizes music preference
-2. **Ready-to-Recommend** (4-6 factors): Can provide initial recommendations, still gathering more info
-3. **Refining** (6+ factors or recommendations made): Explores alternatives, answers detailed questions
-
-**State Management:**
-
-Each turn receives the updated profile from the previous turn via `sessionState.profile`. The profile accumulates information across the conversation:
-
-```typescript
-{
-  livingSpace?: 'apartment' | 'small-house' | 'large-house' | 'farm',
-  activityLevel?: 'couch-companion' | 'playful-moderate' | 'active-explorer' | 'high-energy-athlete',
-  groomingTolerance?: 'minimal' | 'weekly' | 'daily',
-  familyComposition?: 'single' | 'couple' | 'young-kids' | 'older-kids' | 'multi-pet',
-  allergies?: boolean,
-  musicPreference?: 'classical' | 'jazz' | 'rock' | 'metal' | 'pop' | 'country' | 'electronic' | 'none',
-  conversationPhase: 'gathering' | 'ready-to-recommend' | 'refining'
-}
-```
-
-**Breed Matching Algorithm:**
-
-Scores breeds based on profile factors:
-- Music preference: 30 points (2x weight - CRITICAL!)
-- Activity level: 20 points
-- Living space: 15 points
-- Grooming tolerance: 15 points
-- Family composition: 10 points
-- Hard filters: Allergies require hypoallergenic, young kids require good-with-kids
-
-Returns top 5 breeds with match scores (0-100) and reasoning.
-
-**Knowledge Base:**
-
-The agent uses a curated database of 12 breeds with comprehensive trait profiles:
-- Persian, Ragdoll, Siamese, Bengal, Maine Coon, Abyssinian
-- Sphynx, Devon Rex, Russian Blue, Domestic Shorthair, American Shorthair, Scottish Fold
-
-Each breed profile includes: activity levels, grooming needs, apartment suitability, kid/pet compatibility, hypoallergenic status, **music alignment**, temperament, and size.
-
----
-
-### Archetype 9: External Event Integrator
-
-**Characteristics:** Emits events to external systems, blocks waiting for response, timeout handling
-
-**Use Cases:** Human-in-the-loop approval, API callbacks, webhook handlers, external service integration
-
-#### Human Approval Gate
-
-Requests human approval for decisions (mockable for testing).
-
-```typescript
-import { requestHumanApproval } from '@vibe-agent-toolkit/vat-example-cat-agents';
-
-const decision = await requestHumanApproval({
-  title: 'Breeding Permit Review',
-  description: 'Duke Marmalade III x Lady Whiskers',
-  context: { applicationId: '12345', risk: 'low' }
-});
-
-console.log(decision);
-// { status: 'approved', approver: 'human@example.com', timestamp: '...' }
-// OR
-// { status: 'rejected', reason: 'Genetic coefficient too high', timestamp: '...' }
-```
-
-**Mock Mode:** Default returns instant approval. Set `mockable: false` for real HITL integration (Slack, email, etc.).
-
-**Timeout Handling:** Configurable timeout with fallback behavior (default: 24 hours for human timescale).
-
-**Integration Agnostic:** Does not constrain HOW approval is requested (Slack, email, custom UI). Framework adapters implement integration details.
-
----
-
-## Planned Agents (Coming Soon)
-
-### Archetype 4: Agentic Researcher (ReAct)
-
-**Target:** Breed history researcher with tool-calling and iterative reasoning.
-
-**Use Case:** Research cat breed origins using web search + document analysis tools.
-
-### Archetype 5: Function Workflow Orchestrator
-
-**Target:** Breeding approval pipeline with deterministic multi-agent coordination.
-
-**Use Case:** Validate genetics → Generate name → Request approval → Update registry.
-
-### Archetype 6: LLM Intelligent Coordinator
-
-**Target:** Smart submission router with LLM decision-making at checkpoints.
-
-**Use Case:** Route breeding applications based on complexity (auto-approve simple, escalate complex).
-
-### Archetype 7: Function Event Consumer
-
-**Target:** Pedigree file processor triggered by file upload events.
-
-**Use Case:** Process uploaded pedigree documents and update registry.
-
-### Archetype 8: LLM Event Handler
-
-**Target:** Intelligent triage handler with LLM classification.
-
-**Use Case:** Classify incoming submissions and route to appropriate queues.
-
----
-
-## Running Demos
-
-### Photo Analysis Demo
-
-Demonstrates the photo analyzer with actual test fixture images:
-
-```bash
-bun run demo:photos
-```
-
-**What it does:**
-- Loads 4 cat photos + 2 not-cat photos (bear, robot) from `test/fixtures/photos/`
-- Analyzes each photo using **MOCK MODE** (EXIF metadata + filename patterns)
-- Displays extracted characteristics
-- Shows clear warning that it's not analyzing actual pixels
-
-**Mock Mode vs Real Vision API:**
-- **MOCK MODE** (default): Extracts from EXIF metadata and filename patterns. Fast, free, deterministic. Does NOT analyze actual pixels.
-- **REAL MODE** (future): Set `USE_REAL_VISION=true` to call actual vision API (Claude Vision, GPT-4 Vision). Slow, costs money, analyzes actual pixels.
-
-### Runtime Adapter Demos
-
-To see cat agents used across different frameworks:
-
-```bash
-# Vercel AI SDK demo
-cd packages/runtime-vercel-ai-sdk
-bun run demo
-
-# LangChain demo
-cd packages/runtime-langchain
-bun run demo
-
-# OpenAI SDK demo
-cd packages/runtime-openai
-bun run demo
-
-# Claude Agent SDK demo
-cd packages/runtime-claude-agent-sdk
-bun run demo
-```
-
-These demos show the **same cat agents** working across different runtimes, demonstrating portability.
-
----
-
-## Core Schemas
-
-All agents use shared Zod schemas for type safety:
-
-### CatCharacteristics
-
-```typescript
-{
-  physical: {
-    furColor: string,
-    furPattern?: string,
-    eyeColor?: string,
-    breed?: string,
-    size: 'tiny' | 'small' | 'medium' | 'large'
-  },
-  behavioral: {
-    personality: string[],
-    quirks?: string[]
-  },
-  description: string
-}
-```
-
-### Haiku
-
-```typescript
-{
-  line1: string,  // 5 syllables
-  line2: string,  // 7 syllables
-  line3: string   // 5 syllables
-}
-```
-
-### NameSuggestion
-
-```typescript
-{
-  name: string,
-  reasoning: string,
-  alternatives: string[]
-}
-```
-
-### ValidationResult
-
-```typescript
-{
-  status: 'valid' | 'invalid',
-  reason: string,
-  suggestedFixes?: string[]
-}
-```
-
----
-
-## Architecture Highlights
-
-### Multi-Modal Input Convergence
-
-Photo Analyzer and Description Parser both produce `CatCharacteristics` schema. This enables pipelines that accept **either** images **or** text descriptions interchangeably.
-
-```typescript
-// Pipeline works with EITHER input type
-const characteristics = isImage(input)
-  ? await analyzePhoto(input)
-  : await parseDescription(input);
-
-const name = await generateCatName(characteristics);
-const haiku = await generateHaiku(characteristics);
-```
-
-### Feedback Loop Testing
-
-Name Generator → Name Validator creates a realistic feedback loop:
-- Generator creates names without knowledge of rules
-- Validator rejects ~60-70% of names (quirky rules)
-- Forces retry/iteration patterns
-- Tests multi-turn orchestration
-
-### Framework Portability
-
-All agents are **plain TypeScript functions** with no framework dependencies. Runtime adapters translate agents to framework-specific formats:
-
-- **Vercel AI SDK**: Agents → Tools with `execute` functions
-- **LangChain**: Agents → Tools with structured I/O
-- **Claude Agent SDK**: Agents → Agent objects with tool handlers
-- **n8n**: Agents → Custom nodes with visual wiring
-
-**Same agents, different orchestration.**
-
----
-
-## Development
-
-```bash
-# Install dependencies
-bun install
-
-# Build
-bun run build
-
-# Run tests
-bun run test
-
-# Type check
-bun run typecheck
-
-# Lint
-bun run lint
-
-# Run demos
-bun run demo:photos
-```
-
----
-
-## Test Fixtures
-
-### Processing Cat Photos
-
-This package includes git-friendly test images with embedded EXIF metadata for realistic testing.
-
-**Strategy:**
-- Original images: 1-9MB each (too large for git)
-- Processed images: 13-60KB each (git-friendly)
-- Embedded EXIF metadata contains test expectations
-- Mock mode reads EXIF instead of calling vision API
-
-**Processing images:**
-
-```bash
-# From repo root
-cd packages/dev-tools
-bun run process-images ~/Downloads/cat-photos ../../vat-example-cat-agents/test/fixtures/photos/cats
-```
-
-See `@vibe-agent-toolkit/dev-tools` package for the `process-test-images.ts` utility.
-
-**Fixture structure:**
-
-```
-test/fixtures/photos/
-├── cats/          # Valid cat photos (4 images)
-├── not-cats/      # Negative test cases (bear, robot)
-└── cat-like/      # Ambiguous cases (future: stuffed animals, statues)
-```
-
----
-
-## Contributing
-
-This is a reference implementation that evolves with the VAT framework. Each new archetype implementation helps validate the framework design.
-
-**Current priorities:**
-1. Complete remaining LLM analyzer agents (conversational assistant)
-2. Implement agentic researcher (ReAct pattern with tools)
-3. Add workflow orchestrator (multi-agent pipelines)
-4. Demonstrate event-driven patterns (consumers and handlers)
-
-See [CLAUDE.md](CLAUDE.md) for technical navigation details when contributing.
-
----
-
-## Documentation
-
-- **structure.md** - Package organization and conventions
-- **[CLAUDE.md](CLAUDE.md)** - Technical navigation for AI assistants
-
----
-
-## License
-
-MIT