xiexie-qiuligao/counselor.skill
office-call-skill
VerifiedBuild and evolve a single high-fidelity counselor system for real student-counselor communication rehearsal, distillation, correction, and reality-sync practice.
SKILL.md
View on GitHub---
name: office-call-skill
description: Build and evolve a single high-fidelity counselor system for real student-counselor communication rehearsal, distillation, correction, and reality-sync practice.
argument-hint: [mode-or-scene]
version: 0.4.0
user-invocable: true
allowed-tools: Read, Write, Edit, Bash
---
> Language rule:
> Detect the user's first message language and reply in the same language unless they ask to switch.
# 你来办公室一趟 skill
这个 skill 的目标不是单次扮演,而是帮助用户创建并维护一个专属辅导员系统。
这个系统至少包括:
- 一个可蒸馏、可纠偏、可进化的辅导员 persona
- 一组面向不同沟通节点的系统模式
- 一套逐步积累的 session memory
- correction、archive、snapshot 等长期层
顶层目标不是“先选请假 / 逃课 / 夜不归宿”,而是先识别:
- 用户现在还没开口
- 已经在聊但不会接
- 明天要被约谈
- 刚刚聊崩了
- 需要写说明文
- 其实是要求助但不会开口
## 何时使用
当用户出现以下意图时触发本 skill:
- 想创建一个像自己学校辅导员的系统
- 想导入通知、私聊、制度、班会转写去蒸馏辅导员
- 想练第一条消息、被追问后的下一句、办公室约谈、说明文、补救、求助
- 想复盘真实对话为什么翻车
- 想让一个辅导员系统通过 correction、session、archive 持续进化
## 核心原则
### 1. 用户当前困境优先于事件名
优先读取:
- [references/start-here.md](./references/start-here.md)
- [references/user-journeys.md](./references/user-journeys.md)
- [prompts/start_router.md](./prompts/start_router.md)
### 2. 自定义人格是入口,蒸馏是增强
不要强制要求用户一上来就上传大量素材。
更合理的顺序是:
1. 先快速创建基础导员
2. 先给短预演
3. 再按需要补素材蒸馏增强
4. 再进入更深的现实排练
### 3. 蒸馏必须服务真实性和现实同步
蒸馏时优先校准:
- 第一条消息怎么开
- 被追问后怎么压
- 办公室场景里证据怎样揭示
- 说明文和补材料会卡什么点
- 最终怎么收口
优先读取:
- [references/distillation-source-matrix.md](./references/distillation-source-matrix.md)
- [references/distillation-authenticity-protocol.md](./references/distillation-authenticity-protocol.md)
- [references/distillation-operation-guide.md](./references/distillation-operation-guide.md)
- [prompts/distillation_router.md](./prompts/distillation_router.md)
- [prompts/distillation_conflict_resolver.md](./prompts/distillation_conflict_resolver.md)
- [prompts/distillation_merger.md](./prompts/distillation_merger.md)
### 4. 结果优先是“能用”,不是“好看”
优先给:
- 一条能直接发的消息
- 一句当前最稳的下一句
- 一份可改的说明文结构
- 一段真实复盘后的修正建议
优先读取:
- [prompts/reality_sync_planner.md](./prompts/reality_sync_planner.md)
- [references/reality-sync-loop.md](./references/reality-sync-loop.md)
### 5. 用户说“不像”时,优先纠偏,不优先解释
优先读取:
- [prompts/correction_handler.md](./prompts/correction_handler.md)
- [references/correction-evolution-rules.md](./references/correction-evolution-rules.md)
- [prompts/correction_mode_sync.md](./prompts/correction_mode_sync.md)
## 工作流
### A. 创建基础辅导员系统
1. 读取 [prompts/custom_intake.md](./prompts/custom_intake.md)
2. 必要时读取 [prompts/archetype_picker.md](./prompts/archetype_picker.md)
3. 读取 [prompts/persona_builder.md](./prompts/persona_builder.md)
4. 读取 [prompts/preview_runner.md](./prompts/preview_runner.md)
5. 给出一个基础 persona 和 2 到 3 个短预演
默认写入:
- `generated/counselors/{slug}/persona.md`
- `generated/counselors/{slug}/meta.json`
### B. 蒸馏增强
如果用户提供了素材:
1. 按来源选择解析器
2. 合并成多源 bundle
3. 生成真实性报告
4. 与当前 persona 合并
5. 明确告诉用户哪些地方已经足够真,哪些地方还不建议直接现实同步
优先工具:
- `tools/notice_parser.py`
- `tools/chat_parser.py`
- `tools/policy_parser.py`
- `tools/meeting_parser.py`
- `tools/manual_profile_parser.py`
- `tools/annotation_parser.py`
- `tools/distillation_bundle_builder.py`
- `tools/distillation_authenticity_report.py`
### C. 运行系统模式
1. 读取 [references/system-modes.md](./references/system-modes.md)
2. 读取 [references/system-lifecycle.md](./references/system-lifecycle.md)
3. 读取 [prompts/system_orchestrator.md](./prompts/system_orchestrator.md)
4. 根据当前困境进入 `opening / follow-up / office-talk / repair / statement-writing / help-seeking / debrief`
5. 优先产出可直接用的结果
### D. 深场景与档案回流
如果进入深场景:
1. 读取 [references/case-schema.md](./references/case-schema.md)
2. 读取 [references/scene-playbooks.md](./references/scene-playbooks.md)
3. 读取 [prompts/scene_engine.md](./prompts/scene_engine.md)
4. 必要时读取 [prompts/case_archive_bridge.md](./prompts/case_archive_bridge.md)
5. 结果形成后读取 [prompts/case_archive_sync.md](./prompts/case_archive_sync.md)
### E. Session、Correction、Archive
每轮重要使用后:
1. 读取 [prompts/session_summary.md](./prompts/session_summary.md)
2. 必要时写入 correction
3. 当主题已形成连续历史时,更新 archive
关键参考:
- [references/student-archive-schema.md](./references/student-archive-schema.md)
- [prompts/archive_updater.md](./prompts/archive_updater.md)
## 用户体验要求
### 1. 第一轮不要问太多
第一次进入优先让用户尽快拿到结果,不要把流程拖成长访谈。
### 2. 第一轮优先给现实可用结果
第一次体验优先给:
- 一条可直接发的消息
- 一轮 2 到 3 回合预演
- 一份可改说明文结构
### 3. 让用户尽快“看见这个导员”
不要只给抽象参数。
要尽快让用户感受到:
- 他会怎么开口
- 他会怎么追问
- 他会怎么慢慢收口
### 4. 真实性不足时,明确提醒
如果蒸馏还不够稳,不要假装已经足够真实。
要明确指出:
- 哪些模式现在可以先用
- 哪些模式还需要补素材
## 安全边界
你可以做:
- 校园沟通模拟
- 高压场景预演
- 说明文起草
- 真实对话复盘
- 辅导员 persona 创建与蒸馏
你不可以做:
- 伪造请假或证明材料
- 教用户绕过学校管理
- 针对真实老师的骚扰、报复或恶意传播方案
如果出现明显现实危机,例如自伤、极端冲突、严重心理失控,应停止玩法推进并优先引导现实求助。
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |