/ Directory / Playground / wshobson agents
← All servers ↗ GitHub
● Community wshobson ⚡ Instant

wshobson agents

by wshobson · wshobson/agents

83 production-ready Claude Code subagents — backend, frontend, DevOps, data, security — orchestrated by orchestrator-agent.

wshobson/agents is one of the most-used Claude Code subagent collections. 83 specialist agents (backend-architect, frontend-developer, code-reviewer, security-auditor, etc.) plus an orchestrator-agent that delegates among them. Drop into ~/.claude/agents and Claude auto-discovers and routes work.

▶ Watch demo 📦 Install 💡 Use Cases

Why use it

Key features

Live Demo

What it looks like in practice

ready

Install

Author repository · Claude Skills documentation

Use Cases

Real-world ways to use wshobson agents

Build a feature end-to-end with auto-routed specialists

👤 Solo devs building full-stack features ⏱ ~60 min intermediate

When to use: You want backend + frontend + tests + review without context-switching prompts.

Flow
  1. Kick off with orchestrator
    Use orchestrator-agent. Build a CRUD feature for 'invoices' — backend (FastAPI), frontend (Next.js), tests, review.✓ Copied
    → Plan with subtasks routed to specialist agents
  2. Watch each handoff
    Show me each agent's contribution as it goes. Pause for my approval after backend before frontend starts.✓ Copied
    → Per-agent diff with checkpoints
  3. Final security pass
    Run security-auditor on the diff. Anything risky?✓ Copied
    → Security report with findings

Outcome: Full-stack feature shipped with appropriate specialists per layer.

Pitfalls
  • Orchestrator over-delegates trivial tasks — Override with --no-orchestrator for simple changes
Combine with: desktop-commander-mcp

Run a code review with the dedicated code-reviewer agent

👤 Anyone who wants more than 'looks good' on their PR ⏱ ~20 min beginner

When to use: You want a thorough review that doesn't just rubber-stamp.

Flow
  1. Stage the diff
    Run code-reviewer on the current branch's diff vs main. Group findings by severity.✓ Copied
    → Sev-1/2/3 grouped findings
  2. Apply fixes
    Apply the sev-1 findings now. Park sev-2/3 for later.✓ Copied
    → Diff updated; suggestions tracked

Outcome: Higher signal review than 'lgtm'.

Combine with: wshobson-commands-skill

Invoke a specific specialist directly without orchestration

👤 Devs who know exactly which expertise they need ⏱ ~30 min intermediate

When to use: Targeted task; no need for plan-and-route.

Flow
  1. Pick the agent
    Use database-architect agent. Design a schema for multi-tenant orders with audit trail.✓ Copied
    → Schema with rationale
  2. Iterate with same agent
    Stay in database-architect. Now add a hot/cold partition strategy for 100M rows.✓ Copied
    → Refined design

Outcome: Domain-specific output without orchestrator overhead.

Combinations

Pair with other MCPs for X10 leverage

wshobson-agents-skill + wshobson-commands-skill

Slash commands trigger specific agents fast

/review triggers code-reviewer; /audit triggers security-auditor.✓ Copied
wshobson-agents-skill + desktop-commander-mcp

Agents plan, desktop-commander executes

Backend-architect plans; desktop-commander writes the files.✓ Copied

Tools

What this MCP exposes

ToolInputsWhen to callCost
orchestrator-agent high-level goal Multi-domain features Coordinator overhead — multi-step
backend-architect service spec Backend decisions 0
frontend-developer ui spec UI work 0
code-reviewer diff Pre-merge gate 0
security-auditor diff or code Risky changes 0

Cost & Limits

What this costs to run

API quota
N/A — local skill
Tokens per call
High in orchestrator mode (multi-agent coordination)
Monetary
Free (MIT)
Tip
Skip orchestrator for single-domain tasks; call specialist directly

Security

Permissions, secrets, blast radius

Minimum scopes: filesystem-read
Credential storage: None — pure prompts
Data egress: None (skill is local files)

Troubleshooting

Common errors and fixes

Agent not found

Make sure files are in ~/.claude/agents/ with .md extension and proper frontmatter

Verify: claude --list-agents
Orchestrator loops forever

Add explicit termination criteria to your prompt; orchestrator obeys 'stop when X'

Wrong specialist routed

Check frontmatter description fields — orchestrator routes by description match

Alternatives

wshobson agents vs others

AlternativeWhen to use it insteadTradeoff
wshobson/skills (sibling)You want skills (Anthropic skill format) not subagentsDifferent format; same author
obra/superpowersCurated multi-skill bundle with different opinionsSmaller, more opinionated

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills

Continue exploring

Related tools and workflows

Choose another guide for your workflow, or explore a collection.

Browse all topic collections → · See 29 new official additions →