← 返回 Skills 市场
xuanranc

Junyi Client Follow-up

作者 XuanranC · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
115
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install junyi-client-followup
功能描述
儿童成长规划月度跟进工具(客户版)。定时发送复盘问卷、收集家长反馈、格式化整理后转发给规划师。不做评估、不做策略调整、不改报告。覆盖 0-8 岁。触发词:月度跟进、成长复盘、followup、跟进一下孩子、复盘一下、发问卷。不处理:出规划、评估进展、修改报告、医学诊断。
使用说明 (SKILL.md)

junyi-client-followup

月度跟进问卷收集器。定时提醒家长填写复盘问卷,收集反馈后整理转发给规划师(君一)。

Core Rules

  1. 只收集,不评估。 不对家长反馈做任何分析、评分、策略调整。
  2. 不碰报告。 不读取、不修改规划报告文档内容。
  3. 幂等设计。 按 plan_id + 年月去重,同月不重复发送问卷。
  4. 隐私规则:不在群聊中展示孩子信息,memory/ 只写匿名 ID + 状态。
  5. 语气温暖友好,不制造焦虑。

memory/ 写入白名单

只允许写入:child_uid、plan_id、status、state、followup_count、最近跟进日期。 禁止写入:孩子姓名、家长原话、评估细节。

Exit Conditions

  • 找不到 plan_meta 或 status ≠ active → 告知用户需先联系规划师生成规划
  • followup_count ≥ 12 → 提醒:规划已跟进满一年,建议联系规划师评估续期
  • 连续 3 月未回复 → 通知规划师
  • 孩子超龄(0-3 岁超 36 月龄 / 3-8 岁超 8 岁)→ 提醒联系规划师评估新阶段

高风险红线

家长反馈中出现以下情况时,不做常规整理,立即通知规划师:

  • 明显退行(已有能力消失)
  • 极端情绪问题(持续暴怒、自伤倾向)
  • 校园安全事件
  • 家庭重大变故

处理方式:state → flagged,立即发消息给规划师,附上家长原文。

配置

安装时需在 openclaw.json 中配置:

{
  "skills": {
    "entries": {
      "junyi-client-followup": {
        "env": {
          "PLANNER_CONTACT": "规划师联系方式(飞书 open_id 或其他)"
        }
      }
    }
  }
}

Workflow

Mode: send-checkin

触发:cron 每月 1 号 / 手动说"发跟进问卷" 前置:state 必须为 idle

  1. 读取规划文档末尾的 plan_meta
  2. 检查 Exit Conditions
  3. 根据 age 判断年龄段:
    • age_months 存在 → 0-3 岁问卷
    • age_years 存在 → 3-8 岁问卷
  4. 读取 references/client-questionnaire.md 获取对应问卷模板
  5. 填充变量(孩子姓名、月份、年龄)
  6. 发送问卷给家长
  7. 更新 plan_meta:state → sent,current_cycle += 1

Mode: collect-reply

触发:收到家长回复 / 手动说"处理跟进回复" 前置:state 必须为 sent 或 reminded

  1. 读取家长回复原文
  2. 高风险红线检测(读取 references/client-rules.md
    • 命中红线 → state → flagged,立即通知规划师,附原文
    • 未命中 → 继续
  3. 格式化整理(不分析、不评价):
    📋 {孩子姓名} {YYYY年M月} 月度反馈
    ──────────────────
    孩子年龄:{X岁Y个月}
    plan_id:{plan_id}
    跟进次数:第 {N} 次
    ──────────────────
    【家长反馈原文】
    {整理后的反馈,保持原意,分条列出}
    ──────────────────
    ⏰ 请规划师评估后更新规划文档
    
  4. 将整理好的反馈发送给规划师(PLANNER_CONTACT)
  5. 更新 plan_meta:state → idle,last_followup → 当前年月,followup_count += 1
  6. 回复家长:"收到你的反馈,已转给规划师,评估后会给你更新建议 ☺️"

未回复处理

  • 发送后 7 天未回复 → 发一次温和提醒,state → reminded
  • 提醒后仍未回复 → state → skipped,通知规划师"{孩子} 本月未回复"

Read References When Needed

  • 问卷模板:references/client-questionnaire.md
  • 跟进规则:references/client-rules.md
安全使用建议
Before installing, ask the publisher to clarify and (ideally) update the skill metadata and documentation: 1) Explicitly declare required configuration/credentials (PLANNER_CONTACT and any messaging channel tokens) in registry metadata so you know what will be supplied. 2) Resolve the contradiction about "不碰报告" vs reading plan_meta — confirm where plan_meta is stored and ensure reading it does not expose full report content. 3) Confirm exactly what data is written to memory and what is only transmitted transiently; ensure PII (child name, parent messages) is never persisted if that is the intent. 4) Verify the outbound endpoints and delivery mechanism (platform messaging, Feishu, email, etc.) and ensure they are acceptable and secure. 5) Ensure redline forwarding behavior and retention policy are approved by privacy owners (original parent text is forwarded to planners in redline cases). If the publisher cannot or will not clarify these points, treat the skill cautiously — it poses privacy/configuration risks despite appearing functionally coherent.
功能分析
Type: OpenClaw Skill Name: junyi-client-followup Version: 1.0.0 The skill is a monthly follow-up tool for child development planning, designed to collect parent feedback via questionnaires and forward it to a designated planner. It features well-defined safety boundaries, including 'red line' detection for high-risk situations (e.g., developmental regression or safety issues) that triggers immediate human escalation, and strict privacy rules regarding what data can be stored in persistent memory (SKILL.md, references/client-rules.md). No malicious intent, data exfiltration, or execution vulnerabilities were found.
能力评估
Purpose & Capability
The skill claims it will not read or modify planning reports, yet the workflow explicitly says it will "读取规划文档末尾的 plan_meta" (read plan_meta). It also requires a planner contact to forward feedback, but this required configuration is not declared in the registry metadata as an env/credential. These are mismatches between stated purpose/constraints and required actions.
Instruction Scope
Runtime instructions include reading plan_meta, reading parent replies (including original text for redline cases), formatting and forwarding messages to a planner contact, and writing limited memory fields. There are internal contradictions: a privacy whitelist forbids writing child names and parents' original words to memory, but the spec also requires including child names and (for redlines) the parent's original text in messages to the planner. The skill does not specify the messaging endpoints/protocols it will use to send questionnaires or forward replies.
Install Mechanism
Instruction-only skill with no install spec or code files, so nothing is written to disk at install time. This is the lowest-risk install mechanism. However, it does require configuration in openclaw.json which is described in SKILL.md (see environment/proportionality).
Credentials
SKILL.md requires an openclaw.json entry setting PLANNER_CONTACT (planner's contact like Feishu open_id), but the registry metadata lists no required env vars or primary credential. The skill will necessarily need access to messaging/notification capabilities and the planner contact; those requirements are not declared, creating a gap between declared and actual needed credentials/config.
Persistence & Privilege
always is false and there is no install-time persistence. The skill claims limited memory writes (child_uid, plan_id, status, state, followup_count, 最近跟进日期) and explicitly forbids storing PII like child name or parents' original text. That limited, scoped persistence is proportionate — but see instruction_scope concerns about contradictions around forwarding original parent text.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install junyi-client-followup
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /junyi-client-followup 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release: 儿童成长规划月度跟进问卷收集器
元数据
Slug junyi-client-followup
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Junyi Client Follow-up 是什么?

儿童成长规划月度跟进工具(客户版)。定时发送复盘问卷、收集家长反馈、格式化整理后转发给规划师。不做评估、不做策略调整、不改报告。覆盖 0-8 岁。触发词:月度跟进、成长复盘、followup、跟进一下孩子、复盘一下、发问卷。不处理:出规划、评估进展、修改报告、医学诊断。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 115 次。

如何安装 Junyi Client Follow-up?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install junyi-client-followup」即可一键安装,无需额外配置。

Junyi Client Follow-up 是免费的吗?

是的,Junyi Client Follow-up 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Junyi Client Follow-up 支持哪些平台?

Junyi Client Follow-up 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Junyi Client Follow-up?

由 XuanranC(@xuanranc)开发并维护,当前版本 v1.0.0。

💬 留言讨论