菜单
Use when the user wants to install, launch, or build something with the cuga agent framework (pip package `cuga`), or asks "how do I use cuga" / "how do I build an agent with cuga" — routes to the right cuga-harness-kit skill instead of guessing.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when the user wants to govern agent behavior with a cuga policy - block/redirect an intent, add step-by-step playbook guidance, require approval before a tool runs, enhance a tool's description, or reshape agent output format.
Use when the user wants to write Python code that creates or invokes a CugaAgent or CugaSupervisor, e.g. "build an agent with cuga", "how do I call agent.invoke", "set up a multi-agent supervisor".
Use when the user wants a running CugaAgent to load a new capability at runtime via cuga's own skill system, e.g. "add a skill to my agent", "create a .cuga/skills SKILL.md", "how does load_skill work".
Use when the user wants to install cuga, start the demo/web UI, or asks what `cuga start` does or which service/mode to run.
Use when the user wants a cuga agent to ingest, search, or answer questions from documents (PDF/DOCX/XLSX/PPTX/HTML/Markdown/images) - RAG / knowledge base features.
Use when the user wants to give a cuga agent a new capability via a Python function, REST/OpenAPI service, or MCP server, e.g. "add a tool", "connect this API to cuga", "register an MCP server".
| name | cuga-getting-started |
| description | Use when the user wants to install, launch, or build something with the cuga agent framework (pip package `cuga`), or asks "how do I use cuga" / "how do I build an agent with cuga" — routes to the right cuga-harness-kit skill instead of guessing. |
cuga (uv add cuga, github.com/cuga-project/cuga-agent) is an agent orchestration framework: a reasoning engine + pluggable tools (OpenAPI/MCP/LangChain) + a policy system + optional RAG knowledge base + multi-agent supervision, built on LangGraph.
This skill is the entry point. Read the description of the skill that matches what the user is doing before improvising:
| User is trying to... | Use skill |
|---|---|
| Install cuga or start the demo UI | cuga-install-and-launch |
Write Python code that creates/invokes a CugaAgent or CugaSupervisor | cuga-build-agent |
Author a new runtime skill the agent itself can load (.cuga/skills/<name>/SKILL.md) | cuga-build-cuga-skill |
| Register a Python function, OpenAPI spec, or MCP server as a callable tool | cuga-build-tool |
| Add a policy: block an intent, add a playbook, require approval, enhance a tool description, or reshape output | cuga-author-policy |
| Ingest/search documents (RAG) | cuga-knowledge-rag |
| An agent run misbehaved and you need to inspect why | cuga-debug-trajectory |
Don't conflate these — they are unrelated mechanisms that happen to share the SKILL.md filename convention:
.cuga/skills/<name>/SKILL.md) are loaded by a running CugaAgent via a load_skill tool, and are what the agent itself can call at runtime. See cuga-build-cuga-skill when the user wants one of these.cuga's CLI and SDK surface can change between versions. If something in these skills looks stale (a flag, a command, a class name), verify against the installed package before trusting the skill text: python -c "import cuga; print(cuga.__file__)", cuga --help, or the repo at src/cuga/ if working inside a checkout.