design-agent-blueprint

name	design-agent-blueprint
description	Transform structured user requirements into a complete, normalized, and reviewable AgentBlueprint, then save it as the current pending blueprint for confirmation. This skill designs the agent blueprint only and must not create repository files or perform scaffold generation.
triggers	["蓝图","blueprint","设计agent","规划agent","创建agent","设计智能体"]
required_tools	["save_pending_blueprint"]
permissions	[]
input_schema	{"type":"object","required":["requirements"],"properties":{"requirements":{"type":"object","description":"来自 collect_agent_requirements 的结构化需求对象。","properties":{"agent_name":{"type":"string"},"agent_id":{"type":"string"},"role":{"type":"string"},"goal":{"type":"string"},"style_constraints":{"type":"array","items":{"type":"string"}},"required_skills":{"type":"array","items":{"type":"string"}},"required_tools":{"type":"array","items":{"type":"string"}},"user_constraints":{"type":"array","items":{"type":"string"}},"memory_requirements":{"type":"array","items":{"type":"string"}},"workflow_preferences":{"type":"array","items":{"type":"string"}}}},"existing_pending_blueprint":{"type":"object","description":"当前会话里已存在的 pending blueprint，可用于增量修订或局部覆盖。"},"template_hint":{"type":"string","description":"可选的模板偏好，例如 research_agent、fea_agent、rag_agent。"}}}
output_schema	{"type":"object","required":["status","blueprint","missing_fields","conflicts","blueprint_summary","next_action"],"properties":{"status":{"type":"string","enum":["need_more_info","conflict_detected","draft_blueprint"]},"blueprint":{"type":"object","description":"标准化后的 AgentBlueprint，可直接进入确认阶段。","properties":{"agent_id":{"type":"string"},"name":{"type":"string"},"role":{"type":"string"},"goal":{"type":"string"},"style_constraints":{"type":"array","items":{"type":"string"}},"workspace_docs":{"type":"array","items":{"type":"object","properties":{"name":{"type":"string"},"purpose":{"type":"string"},"required":{"type":"boolean"}}}},"skills":{"type":"array","items":{"type":"object","properties":{"skill_name":{"type":"string"},"purpose":{"type":"string"},"triggers":{"type":"array","items":{"type":"string"}},"inputs":{"type":"array","items":{"type":"string"}},"outputs":{"type":"array","items":{"type":"string"}}}}},"tool_plan":{"type":"object","properties":{"reuse_tools":{"type":"array","items":{"type":"object","properties":{"tool_name":{"type":"string"},"purpose":{"type":"string"}}}},"new_tools":{"type":"array","items":{"type":"object","properties":{"tool_name":{"type":"string"},"purpose":{"type":"string"},"reason":{"type":"string"},"io_schema":{"type":"object"},"implementation_hint":{"type":"string"},"risk_level":{"type":"string","enum":["low","medium","high"]}}}}}},"tool_policy":{"type":"object","properties":{"allow_existing_tools":{"type":"boolean"},"allow_local_tools":{"type":"boolean"},"allow_remote_mcp":{"type":"boolean"},"require_approval_for_repo_write":{"type":"boolean"},"require_audit_logging":{"type":"boolean"}}},"memory_namespaces":{"type":"array","items":{"type":"string"}},"workflow_name":{"type":"string"},"create_chat_entry":{"type":"boolean"},"export_agent_factory":{"type":"boolean"},"confirmation_required":{"type":"boolean"}}},"missing_fields":{"type":"array","description":"仍然缺失、无法稳定生成 blueprint 的关键字段。","items":{"type":"string"}},"conflicts":{"type":"array","description":"当前需求中检测到的冲突项。","items":{"type":"string"}},"blueprint_summary":{"type":"string","description":"面向用户确认的简明蓝图摘要。"},"next_action":{"type":"string","enum":["ask_user_for_missing_info","ask_user_to_resolve_conflict","wait_for_user_confirmation"]}}}
decision_logic	[{"if":"requirements 缺少 role、goal、skills 或关键约束，无法形成稳定蓝图","return":"need_more_info"},{"if":"requirements 中存在明显冲突，例如工具限制与能力目标互相矛盾","return":"conflict_detected"},{"if":"信息完整且蓝图内部一致","return":"draft_blueprint"}]
constraints	["必须输出结构化、完整且可确认的 blueprint，不得返回松散自然语言替代结构化对象。","不得省略 workspace_docs、skills、tool_plan、tool_policy、memory_namespaces、workflow_name、create_chat_entry、export_agent_factory 等关键字段。","blueprint 只负责设计，不得在此阶段创建仓库目录、写代码文件或修改 agents 导出。","不得虚构不存在的工具、技能、模板或用户约束。","优先复用现有工具；只有在明确无法满足需求时，才在 tool_plan.new_tools 中声明新本地工具。","若声明新工具，只能提供 implementation_hint 或最小实现思路，不得伪装成已经落地的真实文件。","agent_id 必须归一化为 snake_case；若 requirements 未明确提供且无法稳定推断，应列入 missing_fields。","confirmation_required 必须为 true，且必须明确告知用户只有输入\"确认创建\"后才会进入实际生成阶段。","必须输出 user_message 字段：向用户说明 blueprint 的核心设计思路、包含的功能模块、以及确认后会发生什么，语言要自然、口语化，不要重复 JSON 结构。"]
failure_modes	[{"case":"incomplete_requirements","effect":"无法生成稳定蓝图，后续脚手架生成风险高"},{"case":"conflicting_requirements","effect":"蓝图不一致，进入生成阶段会导致错误实现"},{"case":"invalid_tool_plan","effect":"blueprint 中工具规划不可执行，无法安全交接给生成阶段"}]
fallback_strategy	[{"when":"incomplete_requirements","action":"返回 missing_fields，并给出最少量、高价值的补充问题"},{"when":"conflicting_requirements","action":"返回 conflicts，并要求用户明确取舍或优先级"},{"when":"invalid_tool_plan","action":"收窄工具方案，优先回退为复用已有工具或简化能力边界"}]
tool_policy	{"audit_logging":true,"sandbox_execution":false,"require_approval_for_write":false}
subagent_allowed	false
dependencies	[]
availability_checks	[]
enabled	true
metadata	{"category":"builder","stage":"blueprint_design","produces":"pending_blueprint","consumes":["requirements","existing_pending_blueprint"]}

Use this skill to convert structured requirements into a reviewable AgentBlueprint.

Execution notes:

1. Normalize the requirements into a complete blueprint object instead of returning loose text.
2. Include workspace docs, skill plan, tool plan, tool policy, memory namespaces, workflow name, chat entry creation flag, and export flag.
3. Prefer existing tools first; only declare new local tools when reuse is insufficient.
4. For a new custom tool, include an implementation_hint or minimal runnable design intent in the tool blueprint, but do not create files in this step.
5. Save the coherent blueprint with save_pending_blueprint only after the blueprint is internally consistent.
6. Clearly tell the user that they must input 确认创建 before any repository files are actually written.
7. 【智能功能推断】如果 requirements 中的 required_skills 和 required_tools 为空，你需要从用户的原始需求描述中主动推断需要什么功能。不要因为 skills/tools 为空就生成一个没有功能的 agent。从 agent 的名称、角色和目标来推断它应该具备什么能力。