Skip to content
Skills Library
Personal productivity · Analysis

remember-preference

/remember-preference

在对话中识别可长期复用的用户属性与偏好,写入或更新三层记忆模型。当用户主动提供身高、体重、版型偏好、尺码反馈,或纠正此前信息时使用。不用于记录一次性信息(如当前浏览的商品)或敏感属性。

Licence
MIT
Last repository change
12.11.2026

Assessment and security

95/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content30 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 12.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

在对话中识别可长期复用的用户属性与偏好,写入或更新三层记忆模型。当用户主动提供身高、体重、版型偏好、尺码反馈,或纠正此前信息时使用。不用于记录一次性信息(如当前浏览的商品)或敏感属性。

Original author description

在对话中识别可长期复用的用户属性与偏好,写入或更新三层记忆模型。当用户主动提供身高、体重、版型偏好、尺码反馈,或纠正此前信息时使用。不用于记录一次性信息(如当前浏览的商品)或敏感属性。

How it works

  1. 目的
  2. 什么时候记 / 不记
  3. 写入规则
  4. 输出
  5. 失败处理
Original SKILL.md headings · 84 lines, 237 words
SKILL.md preview
---
name: remember-preference
description: 在对话中识别可长期复用的用户属性与偏好,写入或更新三层记忆模型。当用户主动提供身高、体重、版型偏好、尺码反馈,或纠正此前信息时使用。不用于记录一次性信息(如当前浏览的商品)或敏感属性。
license: MIT
metadata:
  version: "1.0"
  failure_mode: F3 记忆失忆
---

# 记住用户偏好

## 目的

让对话有"复利"——用户说过一次的事,不应该再说第二次。

JD 里的说法是"建立长期、可积累的个性化用户理解"。**"可积累"三个字,落点就是这里。**

## 什么时候记 / 不记

| 信息 | 记? | 层 | 理由 |
|---|---|---|---|
| 身高、体重 | ✅ | L2 | 稳定,下次一定用得上 |
| 版型偏好(紧/宽松) | ✅ | L2 | 稳定 |
| "上次 M 太紧" | ✅ | L3 | 含真实结果,权重最高 |
| 常穿尺码 | ✅ | L2 | 稳定 |
| 预算区间 | ✅ | L2 | 相对稳定 |
| **当前浏览的商品** | ❌ | L1 | 一次性,用完即弃 |
| **三围、健康状况、孕期** | ⚠️ | 不主动记 | 敏感属性。用户主动说的可以当轮使用,不落库 |
| **支付信息、地址** | ❌ | 绝不 | 安全红线 |

**判断标准只有一句:这个信息下次对话还用得上吗?**

## 写入规则

### 规则 1:每条记忆都要有来源和置信度

```json
{
  "field": "height_cm",
  "value": 165,
  "source": "user_stated",   // user_stated | inferred | purchase_history
  "confidence": 0.9,
  "updated_at": "2026-11-10T21:30:00"
}
```

| source | confidence 初始值 | 说明 |
|---|---|---|
| `user_stated` | 0.9 | 用户这轮明确说的 |
| `purchase_history` | 0.8 | 历史订单推断 |
| `inferred` | 0.5 | 模型推的,需再次确认才提升 |

### 规则 2:新信息与旧记忆冲突时,用新的,并提示更新

> 记忆:60kg(三个月前)
> 用户本轮:"我最近胖了十斤"
>
> ✅ 正确:"好的,那按 65kg 帮您重新看一下。要不要我更新一下您的档案?"
> ❌ 错误:继续用 60kg 算
> ❌ 错误:悄悄改成 65kg 不告知

**冲突处理三件套:用新的 → 说明变化 → 征得同意再落库。**

### 规则 3:用户纠正时必须能改

用户说"你记错了,我是 168 不是 165" → 立即更新,并确认"已经改成 168 了"。
**不能因为置信度高就拒绝更新。**

### 规则 4:不要把推断当事实存

如果只是根据用户"我比较瘦"推断出体重,`source` 必须是 `inferred`、`confidence` 0.5,
并且下次使用时要说"您之前提到偏瘦,我按 55kg 左右估算的"——而不是当成确定值。

## 输出

```json
{
  "writes": [
    {"layer": "L2", "field": "height_cm", "value": 165, "source": "user_stated", "confidence": 0.9}
  ],
  "conflicts": [],
  "user_notice": null
}
```

有冲突或更新时,`user_notice` 填入要对用户说的话。

## 失败处理

- 用户表达含糊("我挺高的")→ 记为 `inferred` 的低置信标签,不写具体数值。
- 用户明确说"别记我的信息" → 清空并停止写入,当轮仍可用,但不落库。

Versions

The database has no other skills named remember-preference.

Database snapshot: 13.11.2026 · package version de885beb5691 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 95remember
    Write what a session learned into docs/, the workspace's memory of record (D64, D71). Use near the end of any session where the user stated a preference, corrected you, told you a fact about their business or how they work, or a lesson emerged; also when the user says "remember…
  • 100remember
    Capture a durable instruction, preference, or correction into a project rule (or, rarely, a skill or hook). Use whenever the user says "remember this", "remember that", "always do X", "never do Y", "from now on", "going forward", "make this a rule", "don't forget", "stop doing…
  • 90memstore-remember
    Write what lasts to the team memory (memstore) and forget it again. Use when the user states a preference or corrects you, a decision or procedure is settled, or they ask to remember, note or forget something, and the memstore tool remember_memory is available.
  • 90remember
    Capture, correct, forget, and recall durable user preferences, decisions, and project context in an existing memory store or an explicitly chosen local folder. Use for explicit memory requests and relevant recall.
  • 84mindex-remember
    Save a durable preference, verified fact, project decision, or technical conclusion to MIndex. Use when the user explicitly asks Claude to remember something for future sessions.
  • 95remember-me
    长期记忆与关键事件记忆。当用户需要 AI 记住跨会话、跨年的事(家人、身体、工作、感情、承诺),或问「你还记得…」「我以前说过…」「那年发生了什么」时使用。原文全留、不生成摘要、检索零写入。只记人和事,不存密码与凭据。

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.