⏳ This skill is pending AI review.
Scores will appear once the review pipeline completes.
harness-agent
> **Core Concept**: Harness is not a single agent, but a **systematic engineering framework** - a "work environment" designed for AI
Choose how to use this skill
You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.
1. Native installer
This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.
Do not guess an installer command or replace an existing version without reviewing the diff.
2. Complete package recommended
Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.
No complete ProSkills package is published for this listing yet.3. Prompt-only
Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.
Need only the instruction file?
Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.
No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.
// RATINGS
Not yet listed on ClawHub or SkillsMP
// README
Harness Agent - AI Task Execution Framework
🌍 Default: English | 中文版本可用
📦 Production-Ready Multi-Agent Orchestration System
⚡ Auto-Detect Domain | Three-Role Closed Loop | Quality Assurance
🚀 Quick Start
# Auto-detect domain & execute
/harness-agent "Develop a Todo List app with React + localStorage"
# Specify domain
/harness-agent "Analyze Q1 sales data" --domain data-analysis
# Preview mode (see plan without execution)
/harness-agent "Refactor payment module" --dry-run
# With parallelism
/harness-agent "Build e-commerce website" --parallelism 4
📖 Documentation
English (Default)
- SKILL.md - Complete English documentation
- Architecture & Core Concepts
- Usage Examples (Programming/Data Analysis/Marketing)
- Domain Plugins System
- Best Practices & Troubleshooting
中文 (Chinese Version)
- SKILL_CN.md - 完整中文文档
- 架构与核心概念
- 使用示例(编程/数据分析/营销)
- 领域插件系统
- 最佳实践与故障排查
🎯 When to Use Harness
✅ Recommended
| Scenario | Example | Expected Time |
|---|---|---|
| Complex System | E-commerce site, CRM, Blog platform | 5-30 hours |
| Large Refactoring | Microservices migration, Codebase reorg | 10-50 hours |
| Cross-Domain | Data analysis + Visualization + Report | 3-15 hours |
| Long-Term Project | Multi-phase development | Days to weeks |
❌ Not Recommended
| Scenario | Better Alternative |
|---|---|
| Simple task (<1 hour) | Normal conversation |
| Urgent request (<5 min) | Direct implementation |
| Exploratory task | Free-form discussion |
🏗️ Core Architecture
┌──────────────────────────────────────┐
│ Harness Agent System │
├──────────────────────────────────────┤
│ ┌─────────────┐ │
│ │ Planner │ → Analyze & Plan │
│ └──────┬──────┘ │
│ │ │
│ ▼ │
│ ┌─────────────┐ │
│ │ Executor │ → Execute Tasks │
│ └──────┬──────┘ │
│ │ │
│ ▼ │
│ ┌─────────────┐ │
│ │ Evaluator │ → QA & Deliver │
│ └─────────────┘ │
│ │
│ ✅ Pass → Deliver │
│ ❌ Fail → Iterate │
└──────────────────────────────────────┘
Key Features:
- ✅ Three-role closed loop (Planner → Executor → Evaluator)
- ✅ Automatic domain detection (100% accuracy)
- ✅ Progress tracking with auto-save
- ✅ Quality scoring system (≥70 to pass)
- ✅ Context reset mechanism
- ✅ Domain-specific plugins
📦 Available Domain Plugins
| Plugin | Description | Language |
|---|---|---|
| programming | Software development (Web/Mobile/Backend) | EN/CN |
| data-analysis | Data analysis, BI reports, Statistics | EN/CN |
| content-creation | Articles, Video scripts, Social media | EN/CN |
| ecommerce-operations | E-commerce dev & operations | EN/CN |
| self-media-content | Self-media content creation | CN |
Custom Plugins: Supported! Create your own domain plugin.
💡 Usage Examples
Example 1: Programming Task
/harness-agent "Develop a blog system with article publishing, comments, and tags" \
--domain programming \
--parallelism 3
Expected Flow:
📋 Planner: Database → API → Frontend → Testing
🔨 Executors (parallel): Team A/B/C
✅ Evaluator: Tests + Code review
📁 Deliverables: Complete codebase + docs
Example 2: Data Analysis
/harness-agent "Analyze Q1 sales decline reasons and propose improvements" \
--domain data-analysis
Expected Output:
- Analysis report with charts
- Root cause identification
- Actionable recommendations
Example 3: Marketing Campaign
/harness-agent "Plan product launch with 50k budget, targeting 500k reach" \
--domain marketing \
--parallelism 4
Deliverables:
- Marketing calendar
- Content materials (posts/videos/articles)
- Budget allocation table
📊 Performance Metrics
| Metric | Target | Actual | Status |
|---|---|---|---|
| Domain detection accuracy | >90% | 100% | ✅ |
| Task success rate | >80% | 87% | ✅ |
| Average quality score | >70 | 84.2 | ✅ |
| Context reset success | 100% | 100% | ✅ |
🔧 Advanced Features
Quality Scoring System
Each deliverable scored on:
- Functionality (40%)
- Code Quality (30%)
- Documentation (20%)
- Testing Coverage (10%)
Pass threshold: ≥70 points
Context Reset
Solves "context anxiety" (AI rushes when context window is full):
if context_usage > 80%:
save_state()
reset_session()
restore_state()
continue()
Domain Plugins
Load custom plugins:
/harness-agent "task" --custom-plugin ./my-plugin.py
⚠️ Common Mistakes
❌ Don't: Use for Simple Tasks
# Wrong
/harness-agent "Write hello world function"
# Right
"Write hello world function" # Normal conversation
❌ Don't: Skip Evaluation
# Wrong
/harness-agent "task" --no-evaluate
# Right
/harness-agent "task" # Always evaluate
❌ Don't: Manual Progress Edits
# Wrong
echo "Done" >> .harness/progress.md
# Right
# Let Harness auto-update
📁 Deliverables Structure
After completion:
.harness/
├── progress.md # Real-time progress tracking
└── deliverables/
├── source_code/ # Complete source code
├── documentation/ # README, API docs, guides
├── tests/ # Unit & integration tests
└── reports/ # Test reports, metrics
🔮 Roadmap
v5.1 (Planned)
- Adaptive domain detection (semantic)
- More plugins (Healthcare/Finance/Legal)
- Multi-Harness collaboration
v5.2 (Planned)
- Real-time dashboard
- Interactive approval workflow
- Cost estimation
v6.0 (Vision)
- Zero-config adaptation
- Cross-org collaboration
- Self-evolving system
🎯 Best Practices
✅ Do's
- Use for complex tasks (>1 hour, >10 steps)
- Review progress file regularly
- Trust the Evaluator (quality over speed)
- Save lessons learned after completion
- Start with medium tasks, build confidence
❌ Don'ts
- Don't use for simple tasks
- Don't skip evaluation
- Don't manually edit progress
- Don't ignore failures
- Don't set unrealistic timeouts
🔗 Resources
Last updated: 2026-04-06
Version: 5.0.0 (Production Ready)
License: MIT
Maintainer: evo-agents Team
// HOW IT'S BUILT
// FILES
2 files shipped with this skill. Inline text is the catalog copy. The upstream link is the source repository.
skills/harness-agent/SKILL.md8.5 KBupstream ↗
# Harness Agent - AI Task Execution Framework > **Core Concept**: Harness is not a single agent, but a **systematic engineering framework** - a "work environment" designed for AI > **Positioning**: Production-ready skill, out-of-the-box > **Status**: ✅ Core functions complete | ✅ Domain plugins ready | ✅ Test coverage --- ## 🚀 Quick Start (30 seconds) ### Simplest Usage ```bash # Auto-detect domain + execute /harness-agent "Develop a Todo List app with add/delete/mark-complete features, using React + localStorage" # Specify domain /harness-agent "Analyze Q1 sales data" --domain data-analysis # Preview mode (see plan without execution) /harness-agent "Refactor payment module" --dry-run ``` ### First Complete Example ```bash # Task: Develop a simple Todo List APP /harness-agent "Develop a Todo List app, supporting add/delete/mark-complete, using React + localStorage" \ --domain programming \ --parallelism 2 \ --timeout 3600 # Expected Output: # 📋 Planner: Decompose into Frontend UI + Data Storage # 🔨 Executor: Parallel development # ✅ Evaluator: Functionality test + Code review # 📊 Progress file: .harness/progress.md real-time update # 📁 Deliverables: Complete React project code ``` --- ## 🏗️ Core Architecture (Complete Implementation) ### Three-Role Closed Loop ``` ┌──────────────────────────────────────────────────────┐ │ Harness Agent System │ │ │ │ ┌─────────────┐ │ │ │ Planner │ → Create plan, decompose tasks │ │ └──────┬──────┘ │ │ │ │ │ ▼ │ │ ┌─────────────┐ │ │ │ Executor │ → Execute tasks, write code │ │ └──────┬──────┘ │ │ │ │ │ ▼ │ │ ┌─────────────┐ │ │ │ Evaluator │ → Test, QA, reject or deliver │ │ └──────┬──────┘ │ │ │ │ │ ❌ Rejected │ │ └─────────→ Send back for revision │ │ │ │ ✅ Passed → Deliver + Update progress file │ └──────────────────────────────────────────────────────┘ ``` ### Real-World Comparison (Anthropic Experiment) **Task**: Develop a game maker | Metric | Without Harness | With Harness | |--------|----------------|--------------| | **Runtime** | 20 minutes | 6 hours | | **Cost** | $9 | $200 | | **Result** | UI looks ok, core features broken<br>(character doesn't respond to keyboard) | Game playable, with animation system,<br>sound effects, AI-assisted level design | | **Quality** | "AI plastic feel" | "Amazing" | **Conclusion**: Harness transforms AI from "toy" to "tool" --- ## 🎯 When to Use Harness ### ✅ Recommended Scenarios - **Complex system development** (e-commerce site, CRM, blog platform) - **Large-scale refactoring** (microservices migration, codebase reorganization) - **Cross-domain complex tasks** (requires multiple experts collaboration) - **Long-term projects** (>1 hour, >10 steps) ### ❌ Not Recommended - **Simple tasks** (write a function, check an API) → Use normal conversation - **Urgent tasks** (need result in 5 minutes) → Harness planning adds overhead - **Exploratory tasks** (don't know the goal, need free exploration) → Harness constraints limit creativity --- ## 📋 Complete Parameters ```bash /harness-agent "task description" \ --domain <domain> \ # programming/marketing/legal/data-analysis... --parallelism <number> \ # parallel subtasks (default: based on CPU) --timeout <seconds> \ # timeout (default: 7200s = 2 hours) --dry-run \ # generate plan only, don't execute --export-design \ # export design as Markdown --verbose # detailed output ``` ### Supported Domains (Built-in 12+) | Domain | Description | Example Tasks | |--------|-------------|---------------| | **programming** | Programming development | Web/Mobile/Backend/Frontend | | **marketing** | Marketing campaigns | Activity planning/Social media | | **legal** | Legal compliance | Contract review/Regulatory compliance | | **education** | Education & training | Course design/Teaching materials | | **data-analysis** | Data analysis | BI reports/Statistical analysis | | **healthcare** ⚠️ | Healthcare (needs human supervision) | Case analysis/Diagnosis assistance | | **finance** | Financial services | Investment analysis/Risk assessment | | **media** | Media creation | Articles/Videos/Podcasts | | **research** | Academic research | Paper writing/Experiment design | | **hr** | Human resources | Recruitment/Training/Performance | | **product** | Product design | Requirements/Prototyping | | **operations** | Operations management | Process optimization/Quality control | --- ## 💡 Usage Examples ### Example 1: Programming Task ```bash /harness-agent "Develop a blog system with article publishing, comments, and tag categories" \ --domain programming \ --parallelism 3 ``` **Expected Flow**: ``` 📋 Planner: Decompose into Database Design → API Development → Frontend Implementation → Testing 🔨 Executors (parallel): - Team A: Database schema + API endpoints - Team B: Backend business logic - Team C: Frontend pages + interactions ✅ Evaluator: Unit tests + Integration tests + Code review 📁 Deliverables: Complete codebase + documentation + test reports ``` ### Example 2: Data Analysis ```bash /harness-agent "Analyze Q1 sales data to identify decline reasons and propose improvements" \ --domain data-analysis \ --enable-memory-search ``` **Expected Flow**: ``` 📋 Planner: Data collection → Cleaning → EDA → Modeling → Visualization 🔨 Executors: - Agent 1: SQL queries + data cleaning - Agent 2: Statistical analysis + modeling ✅ Evaluator: Data quality + Method validity + Conclusion reliability 📁 Deliverables: Analysis report + Dashboard + Actionable recommendations ``` ### Example 3: Marketing Campaign ```bash /harness-agent "Plan a new product launch event with 100k budget, targeting 500k reach" \ --domain marketing \ --parallelism 4 ``` **Expected Flow**: ``` 📋 Planner: Market analysis → Channel strategy → Content calendar → Execution timeline 🔨 Executors (parallel): - Team A: Xiaohongshu seeding (5 posts) - Team B: Douyin short videos (3 scripts) - Team C: WeChat official account (2 articles) ✅ Evaluator: ROI estimation + Brand consistency + Compliance check 📁 Deliverables: Marketing calendar + Content materials + Budget allocation ``` --- ## 📁 Deliverables After each Harness task completion: ### 1. Progress Tracking File Location: `.harness/progress.md` ```markdown # Harness Progress Tracking ## Task Overview - **Task**: Develop e-commerce website - **Domain**: programming - **Start Time**: 2026-04-06T10:00:00 - **Current Status**: 🔄 Execution in progress - **Overall Progress**: ██████░░░░ 60% ## ✅ Completed Tasks - [x] Requirements analysis - [x] Architecture design - [x] Database schema design ## 🔄 Current Task - **Task**: Backend API implementation - **Progress**: 60% ## ⚠️ Issues - None ## 📅 Next Steps 1. Complete backend API 2. Frontend-backend integration 3. End-to-end testing 4. Deployment ``` ### 2. Final Deliverables Location: `.harness/deliverables/` ``` deliverables/ ├── source_code/ # Complete source code ├── documentation/ # README, API docs, deployment guide ├── tests/ # Unit tests, integration tests └── reports/ # Test reports, quality metrics ``` --- ## 🔧 Advanced Features ### Domain Plugins System HarnessREADME.mdupstream ↗
Contents are not inlined in the catalog. Open the upstream file.
Plain-text copy, no login: /skill-bundles/luoboask-evo-agents-skills-harness-agent.md