π§ DSH Expert Mode
One agent preset that turns DSH into a "1 Coordinator + 11 Experts" multi-agent team
δΈζ Β· English
---
## β¨ What it does
Install this preset and DSH automatically becomes a "Chief Coordinator" mode:
| Scenario | Behavior |
|----------|----------|
| Receives task | Identifies domain β delegates to the best expert |
| Complex tasks | Dispatches multiple experts in parallel |
| Simple tasks | Coordinator handles directly β no forced delegation |
| Task complete | Experts stay online for follow-up modifications |
No custom prompts to write. No multi-config to maintain. **Just install and use.**
---
## πΌοΈ Demo

Select the "Expert Mode" preset in DSH workspace to use

5 expert subagents working in parallel, with real-time token usage and timing
---
## π§© 11 Experts
| Expert | Tool | Domain |
|--------|------|--------|
| π Data Analyst | `expert_data_analyst` | Data cleaning, statistics, visualization |
| βοΈ Copywriter | `expert_copywriter` | Marketing copy, content creation, rewriting |
| βοΈ Legal Review | `expert_legal_review` | Contract review, legal risk assessment |
| π Product Manager | `expert_product_manager` | Requirements analysis, PRD writing, competitor research |
| π₯οΈ Frontend Dev | `expert_frontend_dev` | Web frontend implementation, component development |
| π¨ UI/UX Design | `expert_uiux_design` | Interface design, interaction patterns, design systems |
| ποΈ Architect | `expert_architect` | System design, tech selection, architecture review |
| π± Social Media | `expert_social_media` | Multi-platform content distribution, account management |
| π Growth Hacker | `expert_growth` | Growth strategy, conversion funnels, A/B testing |
| πΉ Quant Finance | `expert_quant_finance` | Quantitative models, financial analysis, risk control |
| π° Finance | `expert_finance` | Financial analysis, report interpretation, budget planning |
---
## π§ Core Mechanisms
### π Quick Path
The Coordinator **answers directly** without delegating for:
- Single file read/write/edit
- Simple Q&A (no domain expertise needed)
- Casual chat / greetings
- User says "do it yourself" or "no delegation needed"
- Task completable with a single command
### β‘ Fault Recovery
- Expert call timeout/failure β **auto-retry once**
- 2 consecutive failures β inform user, suggest alternative
- Expert output clearly off-topic β **recall and re-guide**
### π Progressive Disclosure
The Coordinator holds a complete expert methodology index and **injects on-demand** β not dumping all expert personas into context at startup.
| Metric | Before | After | Improvement |
|--------|--------|-------|-------------|
| Expert persona tokens | 3850 chars | 533 chars | **-86%** |
| Total prompt tokens | ~2205 | ~1582 | **-28%** |
### π― Anchored Non-degradation
Solves the "trajectory flip" problem caused by system prompt mutations:
| Anchor | Purpose |
|--------|---------|
| **Style Lock** | Maintain default reasoning style regardless of context changes |
| **Evidence Priority** | Every conclusion backed by data/logic |
| **Step-by-step** | Complex tasks decomposed into single steps |
| **Consistency Check** | Verify reasoning style matches previous turn |
| **Anti-drift** | Proactively save state before context window fills |
### ποΈ Five Anchor Constraints
The Coordinator self-checks five anchor points every turn:
| Anchor | Purpose |
|--------|---------|
| **Review** | One-sentence recap of current subtask |
| **Convergence** | Confirm this step advanced the goal |
| **Anti-drift** | 2 turns with no progress β force strategy switch |
| **Collaboration Check** | Before delegation, verify routing is correct |
| **Resource Awareness** | Monitor token usage; auto-simplify if >70% |
### π― Near-distance Guidance
During delegation, the Coordinator fills in the structured guidance template:
```
ββ Near-distance Guidance ββ
Identity: You are [expert role]
Task: {specific task description}
Input: {input data}
Output format: [extracted from methodology index]
Completion criteria: {clear delivery standard}
ββββββββββββββββββββββββββββ
```
Subagents start with clear "who I am, what to do, how to deliver."
### π Expert Communication Protocol
When cross-expert collaboration is needed:
```
[FROM:expert_data_analyst β TO:expert_frontend_dev]
Task: Design frontend data display based on analysis
Data: {Expert A's conclusion summary}
```
### π Cross Review
High-risk tasks (architecture selection, contract review, financial analysis) trigger multi-expert independent review. The Coordinator synthesizes conclusions.
### πΎ Experience Pool
After completing important tasks, experts extract lessons to `.expert-mode/experts/{name}/lessons.md`, automatically injected for similar future tasks.
### π§ Expert Persistence
Experts stay online after task completion. The Coordinator can wake them up for follow-up modifications with full context preserved.
---
## π¦ Installation
### Method 1: dsh plugin add (recommended)
```bash
dsh plugin --profile web add github:Asher-2000/dsh-expert-mode
```
### Method 2: git clone
```bash
mkdir -p ~/.dsh/.agent-presets
git clone https://github.com/Asher-2000/dsh-expert-mode.git ~/.dsh/.agent-presets/expert-mode
```
**Want English?** The repo includes an `expert-mode-en/` preset:
```bash
cp -r ~/.dsh/.agent-presets/expert-mode/expert-mode-en ~/.dsh/.agent-presets/expert-mode-en
```
### Method 3: Manual download
Download from [Releases](https://github.com/Asher-2000/dsh-expert-mode/releases) and place `preset.yml` + `agent.cordis.yml` into `~/.dsh/.agent-presets/expert-mode/`.
---
## π Usage
After installation, select the Expert Mode preset when creating a new session in DSH Web GUI.
---
## β FAQ
**Q: Does Expert Mode consume extra model quota?**
A: Delegating to expert subagents triggers subagent model calls (DSH subagent mechanism), same as the official subagent feature. Simple tasks are handled directly by the Coordinator with no extra calls.
**Q: Can I add my own experts?**
A: Yes. Copy any expert entry in `agent.cordis.yml`, modify the tool name and persona.
**Q: How does it relate to the official standard preset?**
A: Built on top of the official standard preset combination, preserving the full toolset and only adding the expert delegation layer.
---
## π File Structure
```
.
βββ preset.yml # Preset metadata (name + description)
βββ agent.cordis.yml # Coordinator persona + methodology index + expert tools
βββ expert-mode-en/ # English preset
β βββ preset.yml
β βββ agent.cordis.yml
βββ .expert-mode/ # Expert experience pool
β βββ experts/ # Each expert's lessons.md
βββ assets/ # Screenshots
βββ README.md # English
βββ README.zh.md # δΈζ
βββ LICENSE
```
---
## π Changelog
| Version | Changes |
|---------|---------|
| **v0.6.0** | Quick Path + Fault Recovery + bilingual parity |
| **v0.5.0** | Progressive disclosure + anchored non-degradation (tokens -28%) |
| **v0.4.0** | Five anchors + near-distance guidance + cross review + experience pool |
| **v0.3.0** | Expert persistence + communication protocol |
| **v0.2.0** | Basic multi-expert delegation |
---
## π Community
- π’ Featured in **Awesome DSH Plugin**: https://github.com/awesome-dsh-plugin/awesome-dsh-plugin
- π·οΈ GitHub `dsh-plugin` topic: https://github.com/topics/dsh-plugin
---
## π·οΈ Tags
`dsh` `deepseek-harness` `agent-preset` `expert-mode` `multi-agent` `subagent` `ai-agent` `dsh-plugin`
---
## π License
MIT License β see [LICENSE](LICENSE).