dsh-expert-mode

One preset, one full-stack team: your agent becomes a chief coordinator that routes every task to the right expert.

What problem it solves

A single agent preset usually covers one workflow, and cross-domain tasks force you to juggle multiple prompt and tool configs. This agent preset turns your agent into a chief coordinator commanding 16 domain-expert subagents: it identifies the domain and delegates to the best expert, runs several experts in parallel for complex tasks, and handles simple tasks directly without forced delegation. Experts stay online after a task for follow-up modifications.

It builds on the official standard preset, keeps the full toolset, and ships bilingual documentation and presets — no custom prompts to write, no multi-config to maintain, just install and use.

Key features

• 16 domain experts: six full-stack core (frontend / backend / database / architect / DevOps / QA), three security & data (security / data analyst / UI-UX design), and seven business (product / copywriting / legal / social media / growth / quant finance / finance)
• Smart delegation: auto-identifies the task domain and routes it to the best expert
• Fast track: simple tasks are handled directly — no forced delegation
• Five-anchor constraint: a per-turn self-check prevents topic drift
• Cross review: high-risk tasks get independent multi-expert review
• Experience pool: lessons learned are saved and injected next time
• Fault recovery: auto-retry on timeout, strategy switch on failure
• Progressive disclosure: methodology is injected on demand for ~28% token savings

Quick start

Install, then select the “专家模式” / Expert Mode preset in your DSH workspace and just ask:

User: design a user authentication system for me

Coordinator:
  → identifies domain: backend + security
  → delegates Backend Dev: API design, JWT implementation
  → delegates Security: security audit, vulnerability hardening
  → aggregates the full plan

Who it is for

• Users who want one agent spanning code, design, finance, legal, and more
• Teams that want out-of-the-box multi-expert collaboration without maintaining several configs
• Heavy users who want cross-review for high-risk tasks or token savings