agent-avengers
⚠Review·Scanned 2/18/2026
This skill orchestrates multi-agent workflows to decompose tasks, spawn agents, run them, and consolidate results. It contains explicit execution instructions (e.g., os.system('clear'), python3 scripts/consolidate.py) and network/install commands referencing https://github.com/oozoofrog/agent-avengers.git and the AVENGERS_WORKSPACE env var.
from clawhub.ai·v1.0.0·74.9 KB·0 installs
Scanned from 1.0.0 at 445e2a1 · Transparency log ↗
$ vett add clawhub.ai/oozoofrog/agent-avengersReview findings below
Agent Avengers
🦸 All-in-One Multi-Agent Orchestration
한국어 README
Installation
clawhub install agent-avengers
Or clone directly:
git clone https://github.com/oozoofrog/agent-avengers.git ~/.openclaw/workspace/skills/agent-avengers
Usage
Avengers assemble! [complex task description]
Examples
Avengers assemble! Analyze competitors A, B, and C, then create a comparison report
Avengers assemble! Build a full-stack todo app with React frontend and Node.js backend
How It Works
- Decompose — Break down the task into subtasks
- Compose — Assign specialized agents to each subtask
- Execute — Run in parallel or sequential order
- Consolidate — Collect and merge results
- Report — Deliver final output
Agent Types
| Type | Role |
|---|---|
| 🔬 Researcher | Investigation, data collection |
| 🔍 Analyst | Analysis, pattern discovery |
| 🖊️ Writer | Documentation, content creation |
| 💻 Coder | Code implementation |
| ✅ Reviewer | Quality review |
| 🔧 Integrator | Result consolidation |
Agent Modes
| Mode | Description |
|---|---|
| 🔷 Existing | Use registered agents (watson, picasso, etc.) |
| 🔶 Spawned | Create one-off sub-agents on demand |
| 🟣 Multi-Profile | Involve other OpenClaw bot instances |
| 🔷🔶🟣 Hybrid | Combine all modes (recommended) |
Emergent Collaboration Patterns
- 🗳️ Competitive Draft — Multiple agents work independently, best solution wins
- 🎭 Role Rotation — Agents swap roles each round
- ⚔️ Adversarial Collaboration — Creator vs Critic iterations
- 🧬 Evolutionary Selection — Solutions crossbreed and evolve
- 🐝 Swarm Intelligence — Many micro-agents tackle small chunks
- 🔗 Chain Relay — Output → Input chaining
- 💭 Consensus Protocol — Unanimous agreement required
- 🎪 Cross-Domain Jam — Combine different expertise areas
- 🪞 Meta Observer — Agent watches and coaches the team
- ⏰ Temporal Split — Short/mid/long-term parallel approaches
- 🎰 Task Auction — Confidence-based bidding
- 🧠 Shared Memory — Real-time discovery sharing
Configuration
avengers.yaml:
maxAgents: 5
timeoutMinutes: 120
retryCount: 2
defaultModel: sonnet
Scripts
| Script | Description |
|---|---|
scripts/assemble.py | Task decomposition & plan generation |
scripts/execute.py | Generate execution commands |
scripts/monitor.py | Progress monitoring (supports --watch) |
scripts/consolidate.py | Result consolidation |
License
MIT