菜单
Generate an agent across the host AI tools a project uses. Use when the user wants to create, scaffold, or refactor an agent, subagent or specialized role. Not for other artifacts like skills, rules, commands, hooks.
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。
Initialize or refresh the project memory bank. Use when the user wants to set up or regenerate the project's memory files. Not for updating one memory file after it exists or editing a single rule directly.
Guide a project's journey through AIDD, from first setup to shipping a feature. Use when the user says onboard me, where do I start, or what to do next. Not for listing every installed surface.
Design and validate a new SaaS's architecture into an INSTALL.md via Q&A and stack comparison. Use when the user starts a project, chooses a stack, or picks an architecture pattern. Not for editing an existing stack or scaffolding code.
Route a request to generate a context artifact (skill, rule, agent, command, or hook) to its generator when the kind is unnamed. A named kind triggers its generator directly. Not for listing existing artifacts.
Generate a router-based skill across the host AI tools a project uses. Use when the user wants to create, scaffold, or refactor a skill, or turn a workflow into one. Not for other artifacts like rules, agents, commands, hooks.
Generate a coding rule that governs editor and agent behavior across the host AI tools. Use when the user wants to write, add, or refactor a rule, convention, or coding standard. Not for other artifacts like skills, agents, or hooks.
基于 SOC 职业分类
| name | 06-agent-generate |
| description | Generate an agent across the host AI tools a project uses. Use when the user wants to create, scaffold, or refactor an agent, subagent or specialized role. Not for other artifacts like skills, rules, commands, hooks. |
| argument-hint | capture-agent | write-agent | validate |
Write one canonical agent from intent and render it per confirmed host tool, or once as a plugin source.
| # | Action | Role | Input |
|---|---|---|---|
| 01 | capture-agent | Gather the role, propose names, pick the model | user request |
| 02 | write-agent | Render the agent per tool and write | the role |
| 03 | validate | Check each agent file | the files |
Run the actions in order, 01 → 03, and run each action's ## Test before the next.
Before running an action, read its file in actions/, not only the table or assets.
references/agent-authoring.md: the contract (frontmatter, body shape, naming, quality).references/tool-paths.md: per-tool agent path, frontmatter, format conversions, the gate.assets/agent-template.md: agent file scaffold.