agentleFS
hxhx-winhxhx-win/zhuanli-flow/skills/cn-patent-domain-runtime/SKILL.md

cn-patent-domain-runtime

Skill4 starsChanged 3 months agoMITView on GitHub

当用户明确要求生成/起草/写/出中文发明专利、进入完整专利流程,或希望基于既有 `patent/<patent-slug>/state/patent-iteration-state.json` 继续未完成迭代时使用。也可在 `cn-patent-repo-scout` 已确认方向和核心资料后由用户选择进入。本 skill 提供专利领域 Runtime,由宿主 Agent 按确定性状态机调度主线分析、现有技术检索、起草前评审、技术交底、发明人审核、正式稿起草、代理师审稿、DOCX 导出全流程;不直接挑方向(用 `cn-patent-repo-scout`),也不替代单步执行(已有方向只想跑某一步时用对应单 skill:`cn-patent-mainline-analysis` / `cn-patent-prior-art-search` / `cn-patent-formal-drafting` / `cn-patent-attorney-review` / `cn-patent-docx-export`)。触发词:完整专利流程、起草中文发明专利、`patent-iteration-state.json`、专利 Domain Runtime、domain runtime。

The file

Files
SKILL.md

中文发明专利 Domain Runtime

为宿主 Agent 提供完整专利流程的确定性路由、校验和受控状态迁移。Runtime 不直接运行 LLM、创建子 Agent、调用专业工具或代替宿主执行 must_load;具体规则见下方 reference + scripts。

入口边界

  • 不扫描项目;无明确方向先用 cn-patent-repo-scout
  • 只做保护路径设计或特征分层用 cn-patent-mainline-analysis
  • 已选主线只起草正式稿用 cn-patent-formal-drafting

阶段总图

flowchart LR
    s0[step 0 环境/状态] --> s1[step 1 主线] --> s2[step 2 检索]
    s2 --> s3r[step 3 评审] --> s3d[step 3 交底书] --> s3i[step 3 发明人审核] --> s3p[step 3 分流]
    s3i -.rollback.-> s1
    s3i -.rollback.-> s2
    s3p -.handoff.-> hp[专利部接手] --> ga[Gate A 段]
    s3p -.local.-> ga
    ga --> s4[step 4 正式稿] --> s5[step 5 风险] --> s6[step 6 审稿]
    s6 --> gb[Gate B] --> s8[step 8 修订] --> gc[Gate C] --> done[完成]

阶段执行者

阶段执行者派 subagent
1/2/3 评审/3 交底/4/6对应子 skill是 (N0 纪律)
3 发明人审核cn-patent-disclosure-review否(主跑型)
0/Gate A/handoff/分流/5/7/8/9宿主 Agent否

详见 references/agent-tool-mapping.md。

必读 reference (3 份)

  • references/orchestration-philosophy.md:协作哲学(为什么)
  • references/state-machine-reference.md:状态机入口(怎么查)
  • references/agent-tool-mapping.md:N0 派单契约

脚本入口

  • scripts/domain-runtime.py status|validate|transition:统一 Runtime CLI;新调用优先使用此入口
  • scripts/validate-stage.py --state-path X --stage S --mode enter|exit:断言前置/出口
  • scripts/get-next-step.py --state-path X:计算下一步
  • scripts/handoff-package.py / handoff-pickup.py:交接

脚本调用纪律(硬约束)

  • 宿主进入新 stage 前 → 跑 domain-runtime.py validate --mode enter;返回 ok=false 则必须停下补料或询问用户
  • subagent 返回 / stage 完成后 → 跑 domain-runtime.py validate --mode exit;返回 ok=false 则必须补派或追问
  • exit 通过后由宿主调用 domain-runtime.py transition --from-stage S --to-stage T;禁止宿主直接改 state.current_stage
  • Runtime 在锁内复核来源阶段、前置/出口条件和推导目标,并原子写入流程控制字段
  • 不得凭用户原始请求范围(如"只生成交底书")提前结束循环:状态机推进由 Runtime 的 next_action 决定,返回 "completed" 才算终止
  • 按 Runtime 响应的 executor_info 决定下一步动作:type=main_run_skill 必须先 Read must_load 列表里的文件再继续;do_not_merge_user_questions=true 时禁止合并多档 AskUserQuestion
  • step 0 出口附加:把 env-check.json 的 warnings 数组逐条在对话中 echo 给用户(字体缺失、缺工具等),避免下游 docx 阶段才暴露
  • Runtime 只返回任务契约;宿主 Agent 负责加载 Skill、创建子 Agent、执行工具、用户交互和专业结果回写

Gate

  • Gate A:起草前决策完成 + 用户确认进起草
  • Gate B:审稿意见 + 用户反馈双方就位
  • Gate C:质量检查通过 + 用户授权交付

状态与脚本

  • 状态文件:patent/<patent-slug>/state/patent-iteration-state.json
  • 模板:assets/patent-iteration-state.template.json
  • 历史兼容(旧 state 缺新字段):scripts/new-iteration-state.py --migrate-from <old.json>

子 skill 引用

  • 前置可选:cn-patent-repo-scout
  • 必要前置:cn-patent-mainline-analysis / cn-patent-prior-art-search / cn-patent-disclosure-draft / cn-patent-disclosure-review / cn-patent-formal-drafting / cn-patent-attorney-review
  • 关联可选:cn-patent-docx-export(Gate C 通过后导出)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.