agentlisted
Install: claude install-skill x-mesh/xm
<Purpose>
Two layers of agent capability on top of Claude Code's native Agent tool:
1. **Primitives** — Reusable patterns (fan-out, delegate, broadcast) for direct agent control
2. **Autonomous Behaviors** — Goal-driven agent activities (research, solve, consensus, swarm) where agents self-direct, discover, and collaborate via stigmergy
Key distinction from x-op: x-op = leader-controlled strategies. x-agent = agent-autonomous execution.
</Purpose>
<Use_When>
- User wants to send a task to multiple agents in parallel
- User says "fan-out", "delegate", "broadcast", "ask N agents"
- User wants agents to explore/research autonomously ("research this", "investigate freely", "find out about")
- User wants agents to solve a problem without step-by-step direction ("figure out why", "fix this somehow")
- Other xm skills need agent primitives
- User wants direct agent control without a full strategy
</Use_When>
<Do_Not_Use_When>
- User wants a leader-controlled strategy with fixed phases (use /xm:op instead)
- Single simple task (just use Agent tool directly)
- User says specific strategy names like "refine", "tournament", "debate" (use /xm:op)
</Do_Not_Use_When>
# x-agent — Agent Primitives & Autonomous Behaviors
Structured agent primitives and autonomous behaviors on top of the Claude Code native Agent tool.
No external dependencies. Works with Claude Code alone.
```
x-agent
├── Primitives fan-out, delegate, broadcast, status
├── Autonomous research, solve, co