菜单
Add a new project-specific agent mid-project, with Chief AI Officer drafting the definition and justification, and the user approving. Use when project scope expands (new domain emerges, a parallel track is needed, or critical expertise is missing) during Phase 2.
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
Add a new MCP server to the project mid-project. Wraps the capability-management procedure — external MCP search, ADR draft, .mcp.json update, user install instructions, and capability-log entry. User approval is mandatory regardless of involvement level. Use when an authorized agent identifies a missing external integration.
Add a new project-specific skill mid-project. Wraps the capability-extension procedure — reuse audit, skill-creator invocation, capability-log entry, user approval, and reload reminder. Use when an authorized agent identifies a missing helper-skill capability during Phase 1 STEP 2/3 or anywhere in Phase 2.
Re-activate a Phase 1 agent from STANDBY because a requirement, design, roadmap, team-composition, or documentation drift has been detected. The user invokes this with the target agent name and a brief reason. Use when Phase 2 cannot proceed without a Phase 1 decision.
Produce a production hotfix — branch from main, fix, change-review back into main, then merge the fix back into develop. Use when a bug in the released version (main) must be patched without waiting for the normal develop→main cycle.
Initialize a new project with a structured, multi-agent workflow — discovery interview, operating preferences, technical design, agent team formation, roadmap construction, documentation consolidation, and a final user approval gate. Use when the user starts a fresh project and wants a rigorous setup before any code is written.
Render the active milestone's tasks as a kanban board grouped by lane (backlog / todo / in-progress / in-review / done). Forge-aware — GitHub mode shows the Projects v2 URL alongside ASCII fallback; other forges render ASCII only. WIP overflow is flagged but not blocked. Use whenever the user asks "what's where" or wants a quick board view.
| name | add-agent |
| description | Add a new project-specific agent mid-project, with Chief AI Officer drafting the definition and justification, and the user approving. Use when project scope expands (new domain emerges, a parallel track is needed, or critical expertise is missing) during Phase 2. |
You are the Conductor. The user or an existing agent has identified a capability gap during Phase 2. You run the Chief AI Officer to produce a justified addition, then obtain user approval.
User invokes /atelier:add-agent <proposed-title> <reason>.
<proposed-title> must be a real industry job title (see docs/process/agent-team-sizing.md).<reason> cites the capability gap.Provide:
docs/agents/team-composition.md.Chief AI Officer produces:
<user-project>/.claude/agents/<kebab-title>.md with persona, inputs, outputs, collaboration, reference docs.docs/agents/team-composition.md with the new agent and justification row.docs/flows/agent-document-map.md showing how the new agent fits.Before presenting to user:
Show the user:
Ask:
"Approve adding
<title>to the team?"
docs/agents/team-composition.md under "Considered and declined".Backend Engineer (Billing) alongside existing Backend Engineer requires proof of orthogonality, critical expertise, or parallel-track necessity.docs/process/agent-team-sizing.mddocs/agents/team-composition.mddocs/flows/agent-document-map.md