Chapter 22: wshobson Marketplace Integration (Entry #227)

Status: Production-Validated (Jan 1, 2026) Difficulty: Beginner Time: 15 minutes ROI: 273 pre-built components (67 plugins, 99 agents, 107 skills)

Overview

The wshobson/agents marketplace provides a curated collection of Claude Code components that integrate seamlessly with your existing setup.

What’s Included

Component Type	Count	Examples
Plugins	67	backend-development, database-design, llm-application-dev
Agents	99	backend-architect, ai-engineer, database-optimizer
Skills	107	Various workflow and domain skills

Quick Setup (15 min)

Step 1: Clone the Marketplace

cd ~/.claude
git clone https://github.com/wshobson/agents.git wshobson-agents

Step 2: Link Plugins

# Create plugins directory if needed
mkdir -p ~/.claude/plugins

# Link desired plugins
ln -s ~/.claude/wshobson-agents/plugins/backend-development ~/.claude/plugins/
ln -s ~/.claude/wshobson-agents/plugins/database-design ~/.claude/plugins/
ln -s ~/.claude/wshobson-agents/plugins/llm-application-dev ~/.claude/plugins/
ln -s ~/.claude/wshobson-agents/plugins/observability-monitoring ~/.claude/plugins/

Step 3: Verify Integration

# List available agents
ls ~/.claude/plugins/*/agents/

# Check agent count
find ~/.claude/plugins -name "*.md" -path "*/agents/*" | wc -l

Key Agents

Backend Development

Agent	Purpose	Use When
`backend-architect`	API design, microservices	Creating new services
`graphql-architect`	GraphQL federation, performance	GraphQL APIs
`tdd-orchestrator`	Test-driven development	Writing tests first
`temporal-python-pro`	Workflow orchestration	Long-running processes

Database Design

Agent	Purpose	Use When
`database-architect`	Schema modeling, tech selection	New databases
`sql-pro`	Query optimization, performance	SQL tuning

LLM Application Development

Agent	Purpose	Use When
`ai-engineer`	RAG systems, embeddings, agents	AI features
`prompt-engineer`	Prompt optimization, chain-of-thought	Improving AI
`vector-database-engineer`	pgvector, similarity search	Semantic search

Observability & Monitoring

Agent	Purpose	Use When
`database-optimizer`	Query performance, indexing	Slow queries
`network-engineer`	Cloud networking, security	Network issues
`observability-engineer`	Metrics, tracing, dashboards	Monitoring setup
`performance-engineer`	Load testing, optimization	Performance issues

Integration with Pre-prompt Hook

Add Marketplace Agents to Triggers

Update AUTOMATIC-TOOL-TRIGGERS.md:

# wshobson Marketplace Agents (Entry #227)
RAG/embeddings/prompt engineering/semantic search/vector database:
  → Use llm-application-development-skill
  → Task(subagent_type: "ai-engineer") for complex implementations
  NEVER: Build RAG without skill patterns

API design/REST/GraphQL/microservices/OpenAPI:
  → Task(subagent_type: "backend-architect")
  NEVER: Design APIs without backend-architect guidance

SQL optimization/query performance/database tuning/EXPLAIN ANALYZE:
  → Task(subagent_type: "database-optimizer")
  NEVER: Optimize queries without database-optimizer

observability/monitoring/tracing/Grafana/Prometheus/SLO:
  → Task(subagent_type: "observability-engineer")
  NEVER: Set up monitoring without observability-engineer

Three-Tier Model Strategy

wshobson agents support Claude’s model selection:

Tier	Model	Use For	Agent Examples
Opus	Most capable	Complex architecture	backend-architect, ai-engineer
Sonnet	Balanced	Standard tasks	(default inheritance)
Haiku	Fast/cheap	Quick lookups	simple queries

# In agent Task() call:
Task(
  subagent_type: "backend-architect",
  model: "opus"  # For complex architectural decisions
)

Hybrid Skills Pattern

Combine wshobson agents with your custom skills:

Example: LLM Application Development Skill

---
name: llm-application-development-skill
description: |
  Hybrid skill combining wshobson ai-engineer patterns with project-specific
  PostgreSQL/pgvector implementation. Use for RAG, embeddings, semantic search.
---

## When to Use

- Building RAG pipelines
- Implementing semantic search
- Adding embeddings to PostgreSQL
- Prompt engineering optimization

## wshobson Agent Integration

```bash
# For complex implementations, delegate to:
Task(subagent_type: "ai-engineer")

# For prompt optimization:
Task(subagent_type: "prompt-engineer")

Project-Specific Patterns

pgvector with HNSW indexes
Hebrew embeddings (text-multilingual-embedding-002)
Production caching strategies ```

Validation

Check Integration (2 min)

# Verify plugins linked
ls -la ~/.claude/plugins/

# Count available agents
find ~/.claude/plugins -name "*.md" -path "*/agents/*" | wc -l
# Expected: 15-20 agents (from 4 plugins)

# Test agent availability in session
# Start Claude Code and ask:
# "What agents are available for backend development?"

Test Agent Routing

Start fresh session and ask:

I need to design a REST API for user authentication

Expected:

✅ backend-architect agent suggested or used
✅ Response includes API design patterns
✅ References REST/microservices best practices

Benefits

Time Savings

Task	Without Marketplace	With Marketplace	Savings
Design new API	2-4 hours	30 min	75%
Set up RAG	4-8 hours	1 hour	85%
Query optimization	1-2 hours	15 min	85%
Monitoring setup	4-6 hours	1 hour	80%

Quality Improvements

Consistent patterns: Agents follow industry best practices
Fewer mistakes: Pre-validated architectural patterns
Better documentation: Agents explain their reasoning

Maintenance

Weekly Update (2 min)

cd ~/.claude/wshobson-agents
git pull origin main

Add New Plugins

# Check available plugins
ls ~/.claude/wshobson-agents/plugins/

# Link new plugin
ln -s ~/.claude/wshobson-agents/plugins/NEW_PLUGIN ~/.claude/plugins/

Troubleshooting

Issue: Agent not found

Check symlinks:

ls -la ~/.claude/plugins/
# Should show -> links to wshobson-agents/

Fix broken links:

rm ~/.claude/plugins/broken-plugin
ln -s ~/.claude/wshobson-agents/plugins/correct-plugin ~/.claude/plugins/

Issue: Agent not triggered

Add to AUTOMATIC-TOOL-TRIGGERS.md:

your-keyword:
  → Task(subagent_type: "agent-name")

Chapter 21: Pre-prompt Optimization (keyword patterns)
Chapter 20: Skills Filtering (agent routing)
Chapter 6: MCP Integration (tool ecosystem)

Implementation Time: 15 minutes Marketplace: https://github.com/wshobson/agents Evidence: Production production (4 plugins, 15+ agents active) Last Updated: 2026-01-05