菜单
Create, update, run, resume, or continue agent workflows. Use when the user asks to create or fix workflow JSON, execute an existing workflow, run a workflow with a prompt, resume a workflow run, or continue the next node in a waiting multi-step 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
Reverse-engineer an existing software project so the user can understand the code logic, architecture, workflows, design decisions, and potential issues. Use this when the user provides a GitHub repo, local codebase, zip file, or existing project and wants architecture understanding, Mermaid diagrams, remake guidance in another language, or expert design/performance evaluation.
Perform deep research from a research requirement or instruction file and produce research outputs in the same folder, including Markdown plus an index.html visual report with Mermaid diagrams when useful.
Open and operate the local Skill Pilot WebUI through the browser for demos, walkthroughs, settings changes, task execution, and terminal-based education flows. Use when the user wants the agent to drive Skill Pilot as a human would in the browser, open a specific WebUI feature, or jump directly into shell or agent terminal views.
Use this skill when acting as a Black-Box Code Agent Supervisor. The supervisor uses coding agent CLIs such as Claude Code, Codex, OpenCode, Gemini CLI, or other AI coding agents through tmux terminal windows to finish a coding task or project. The supervisor works like a real developer using AI agents: giving instructions, checking the running result, using browser/curl/logs, finding issues, and sending feedback back to the coding agent. The supervisor does not directly inspect source code, write code, or write code-level tests.
Manage the full lifecycle of a Vibe Coding project — create, refine, brainstorm, initialize, plan, implement, review, test, deploy, update, and fix issues. Use when the user wants to start a new project from a prompt, evolve its requirements, plan or write code, validate or deploy it, or apply updates and bug fixes against an existing project.
Create interactive courses and assignments for students in the project's markdown format. Use when asked to create a new course, educational module, or assignment.
| name | agent-workflow |
| description | Create, update, run, resume, or continue agent workflows. Use when the user asks to create or fix workflow JSON, execute an existing workflow, run a workflow with a prompt, resume a workflow run, or continue the next node in a waiting multi-step workflow. |
Create, update, run, and continue agent workflows while keeping workflow files valid and terminal execution human-in-the-loop aware.
Workflow JSON keeps runnable nodes as type: "agent". Agent node data uses data.subagent for a Skill Pilot subagent from core/subagents/*/*.md; do not use data.skill for workflow nodes.
As an expert in your assigned roles, you must announce your actions before performing them using the following format:
As a {Role, and Role-XYZ if have more roles}, I will {action description}
This communication pattern ensures transparency and allows for human-in-the-loop oversight at key decision points.
Follow these steps in order:
references/create-update-workflow.md.references/execute-workflow-action.md.references/continue-workflow-action.md.Open the one reference file that matches the selected case. Load additional references only when that case explicitly requires them.
Follow the selected reference file exactly. Keep generated workflow files under core/workflows/ by default, and keep any temporary or intermediate helper files under .skillpilot/temp/ unless the user explicitly requested another project-local path.
Run the validation or execution checks required by the selected case. Report the resulting file path, command, output summary, or blocker in concise language.