summary-worldofai-claude-code-subagents

Claude Code Sub-Agents

Author / channel: WorldofAI Format: video Source: Original Published: 2025-07-29

Summary

Anthropic shipped sub-agents in Claude Code: project- or user-scoped specialist agents with their own system prompt, tool whitelist, and dedicated context window, invokable from the main session. WorldofAI walks the /agents UI, demos a UX-optimizer + UI agent running in parallel on a signup page, and points at the community marketplace at claudecodeagents.com.

Key Points

• Sub-agent definition — each sub-agent has: a specific expertise area, an isolated context window (preserves main-thread context), an explicit tool/MCP whitelist, and a custom system prompt. Configurable at project level or user level.
• Why it matters — solves three Claude Code pain points at once: context-window pollution from off-task work, tool-call hallucination, and the lack of a clean delegation mechanism.
• Setup flow — /agents → New → choose level → describe (recommended: let Claude generate, then customize) → select tools → assign color tag.
• Marketplace — claudecodeagents.com hosts community sub-agents (e.g., UX-optimizer, Delegator). The Delegator demo’d here researches a task, writes specs, and dispatches background agents — an in-terminal AI project manager.
• Parallelism — multiple sub-agents can work in parallel on the same project (UX-optimizer + UI-designer running concurrently on a signup flow).

Connected Pages

• claude-code — sub-agents are now a first-class Claude Code feature alongside /loop and /ultra-plan
• multi-agent-orchestration — sub-agents are Anthropic’s first-party answer to this pattern
• skills-sh — adjacent ecosystem (skills vs sub-agents distinction)
• WorldofAI — author hub