协作头脑风暴
brainstorming
在实现前澄清需求、探索设计选择,并形成清晰的开发方向。
swarms【开发与工程】Route Swarms users to the right single-agent, CLI, workflow, or tool/MCP guidance.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this repo skill for the swarms Python package when the task is about building,
routing, loading, or debugging Swarms agents and orchestrations.
Agent, go to sub-skills/single-agent/.swarms CLI, YAML, or markdown loaders, go to sub-skills/cli-loaders/.sub-skills/multi-agent-workflows/.sub-skills/tools-mcp/.pip install swarms
swarms --help
python -c "from swarms import Agent; print(Agent.__name__)"
WORKSPACE_DIR controls where agent memory and workspace files are stored.SWARMS_API_KEY is required for marketplace fetch/publish workflows.OPENAI_API_KEY, ANTHROPIC_API_KEY, GROQ_API_KEY, and similar provider keys are required only for live model-backed runs.graphviz and rustworkx are optional for graph-workflow variety.references/overview.md for the package map and route selection notes.references/troubleshooting.md for cross-cutting install, import, key, and backend issues.scripts/smoke.sh for a quick offline import and CLI check.Choose the narrowest sub-skill that owns the user-facing workflow. Use the root only when the request spans multiple families or is about the package as a whole.
single-agent owns one-agent configuration and its direct support objects.cli-loaders owns command-line entry points and config/file loading.multi-agent-workflows owns swarm orchestration and routing patterns.tools-mcp owns tool schema helpers and MCP transport/auth workflows.