一键导入
sparkgen-agent
Add, modify, remove, list, or show agents in the workflow
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。
菜单
Add, modify, remove, list, or show agents in the workflow
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
Generate a new SparkGen-AWS project from the cookiecutter template
Develop and modify the SparkGen-AWS cookiecutter template — variables, hooks, files
Run cookiecutter matrix tests to verify template generates correctly across all variable combinations
Send a chat message to the running agent server and display the response
Switch LLM providers, deployment modes, and manage environment configuration
Diagnose issues with Ollama, Docker, AWS, endpoints, guardrails, RAG, or general health
| name | sparkgen-agent |
| description | Add, modify, remove, list, or show agents in the workflow |
| user_invokable | true |
| auto_invokable | true |
| auto_invoke_hint | Invoke when the user discusses agents, agent routing, or handoffs |
| arguments | <add|modify|remove|list|show> [agent-name] |
Manage agents defined in config/ai_workflow.yaml.
Before any action:
config/ai_workflow.yaml — parse the agents: section to get current agent listls prompts/ls contexts/ls guardrails/agents/ 2>/dev/null/sparkgen-agent list)Parse config/ai_workflow.yaml and display a table:
| Name | Role | Tools | RAG Mode | Guardrail Sets |
For each agent, show its key configuration.
/sparkgen-agent show <name>)Display full agent config from workflow YAML including:
handoffs: section)/sparkgen-agent add <name>)Create all required files for a new agent:
prompts/<name>.md with a system prompt templatecontexts/<name>_guidelines.md if the agent needs specific contextguardrails/agents/<name>.md for agent-specific rulesagents: section in config/ai_workflow.yaml:
- name: <name>
role: "<role description>"
description: "<what this agent does>"
tools: [<tool_list>]
prompt:
prompt_file: prompts/<name>.md
context_files: [contexts/platform_context.md]
variables:
agent_name: "<Name>"
prompt_suffix: ""
rag_context_template: ""
guardrails:
use_sets: [platform_defaults]
rag:
enabled: false
mode: standard
knowledge_bases: []
inject_context: false
handoffs: section:
- from: main_agent
to: <name>
condition: "intent == '<name_intent>'"
single_agent, switch to router_managermake validate to verify the workflow loads correctly/sparkgen-agent modify <name>)make validate/sparkgen-agent remove <name>)agents: in workflow YAMLhandoffs:prompts/<name>.md and contexts/<name>_guidelines.mdsingle_agentmake validatemain_agent is the entry point and should not be removed