memory-distill SKILL.md skillA

memory-distill from kendrick-stein/mcislab_deepread is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add kendrick-stein/mcislab_deepread/memory-distill@git:20260723.fb64ba4

Pins git:20260723.fb64ba4 in mdr.lock. Pin the fingerprint instead.

Audit grade
A, 17 of 17 checks passed
Kind
SKILL.md skill
Repository
kendrick-stein/mcislab_deepread
Path
skills/5-evolution/memory-distill/SKILL.md
Versions recorded
2
Latest version
git:20260723.fb64ba4, committed Jul 23, 2026
First version recorded
Apr 28, 2026
Size
11,721 bytes
Repository stars
12
License
none declared

当积累了多天工作日志、或 Supervisor 说"整理记忆""蒸馏"时,从日志中提取 pattern 和 insight 到记忆库。也可被 autoresearch 在合适时机自动调用.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

## Purpose

memory-distill 是 MindFlow 记忆演化体系的基础技能。它定期扫描 `Workbench/logs/` 中的原始工作日志,从中提取跨日期重复出现的 pattern 和意外发现,并将有价值的 observation 提升为结构化的记忆条目,写入 `Workbench/memory/patterns.md` 和 `Workbench/memory/insights.md`。

该技能实现了 `references/memory-protocol.md` 中定义的 Insight Promotion Hierarchy 的底层两级跃迁:从 Level 0(Raw Log)提升至 Level 1(Pattern),再进一步触发 Level 1 → Level 2(Provisional Insight)和 Level 2 → Level 3(Validated Insight)的升级。它将"散落的日常观察"转化为"可复用的研究经验",是 MindFlow 自我进化机制的核心入口。

输入为时间范围内的日志文件(`Workbench/logs/YYYY-MM-DD.md`),输出为更新后的记忆文件和 changelog 条目。

## Steps

### Step 1:收集日志

1. 解析 `period` 参数,确定起止日期。若未提供,默认为今天往前 7 天(含今天)。
2. 用 Glob 列出 `Workbench/logs/YYYY-MM-DD.md` 格式的所有日志文件。
3. 根据文件名中的日期过滤,保留落在 `period` 范围内的文件。
4. 用 Read 逐一读取全部匹配的日志文件,记录每个文件的日期和内容。
5. 若范围内无任何日志文件,输出提示"指定时间段内无日志,跳过蒸馏",终止执行。

### Step 2:提取候选 Pattern

通读所有收集到的日志内容,重点扫描每条 log entry 的 `observation` 字段及其他叙述性文字,寻找以下三类候选 pattern:

1. **跨日期重复观察**:同一现象、规律或结论在两个或更多不同日期的日志中均有提及。即使措辞不同,只要语义相似,均视为同一 pattern 的多次出现。
2. **意外发现(anomaly)**:某个结果或行为与 `DomainMaps/` 中记录的已有知识相悖,或日志中明确标注为"出乎意料"、"与预期不符"的观察。
3. **关联线索(correlation clue)**:日志中提示两篇论文、两个实验或两个概念之间存在潜在联系,且该联系尚未在任何记忆文件中被明确记录。

对每个候选 pattern,记录:
- 核心 observation 的一句话概括
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add kendrick-stein/mcislab_deepread/memory-distill@sha256:926b290f5344e3d5

The line at the top of this page pins the label git:20260723.fb64ba4, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_oxdl5jby37vjp6n7.svg)](https://markdownregistry.com/a/art_oxdl5jby37vjp6n7)

3 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 2 versions of this file, between Apr 28, 2026 and Jul 23, 2026, about one every 86 days.

versioncommittedcommitsizeaudit
git:20260723.fb64ba4 latest2026-07-23 fb64ba4 11,721 BA view · diff
git:20260428.bf023ec2026-04-28 bf023ec 9,932 BA view

Audit of the latest version

A  17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (11721 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

kendrick-stein/mcislab_deepread · 12 stars · license none declared · pushed 2026-10-19 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_oxdl5jby37vjp6n7
GET https://markdownregistry.com/api/v1/resolve?ref=kendrick-stein/mcislab_deepread/memory-distill
GET https://markdownregistry.com/api/v1/blob/926b290f5344e3d5579550780df6c9db4105f961a5a834b6b9537d0a6952dc2c

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

More from kendrick-stein/mcislab_deepread

lark-approval SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-approval/SKILL.md · 飞书审批 API:审批实例、审批任务管理。
v1.0.0 · audit A · 12 stars
lark-attendance SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-attendance/SKILL.md · 飞书考勤打卡:查询自己的考勤打卡记录
v1.0.0 · audit A · 12 stars
lark-base SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-base/SKILL.md · 当需要用 lark-cli 操作飞书多维表格(Base)时调用:适用于建表、字段管理、记录读写、视图配置、历史查询,以及角色/表单/仪表盘管理/工作流;也适用于把旧的 +table / +field / +record…
v1.2.0 · audit A · 12 stars
lark-calendar SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-calendar/SKILL.md · 飞书日历(calendar):提供日历与日程(会议)的全面管理能力。核心场景包括:查看/搜索日程、创建/更新日程、管理参会人、查询忙闲状态及推荐空闲时段、查询/搜索与预定会议室。注意:涉及【预约日程/会议】或【查询/预定会议室】时,必须先读…
v1.0.0 · audit A · 12 stars
lark-contact SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-contact/SKILL.md · 飞书 / Lark 通讯录,用于按姓名 / 邮箱把员工解析成 open_id,以及按 open_id 反查员工的姓名 / 部门 / 邮箱 / 联系方式。当用户说出某人姓名而下一步需要发消息 / 加群 / 排日程时,先用本 skill…
v1.0.0 · audit A · 12 stars
lark-doc SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-doc/SKILL.md · 飞书云文档 / Docx / 知识库 Wiki 文档(v2):创建、打开、读取、获取、查看、总结、整理、改写、翻译、审阅和编辑飞书文档内容。当用户给出飞书文档…
git:20260525.97d682b · audit A · 12 stars
lark-drive SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-drive/SKILL.md · 飞书云空间:管理云空间中的文件和文件夹。上传和下载文件、创建文件夹、复制/移动/删除文件、查看文件元数据、管理文档评论、管理文档权限、订阅用户评论变更事件、修改文件标题(docx、sheet、bitable、file、folder、wiki…
v1.0.0 · audit A · 12 stars
lark-event SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-event/SKILL.md · Lark/Feishu real-time event listening / subscribing / consuming: stream events as NDJSON via `lark-cli event consume…
v1.0.0 · audit A · 12 stars
lark-im SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-im/SKILL.md · 飞书即时通讯:收发消息和管理群聊。发送和回复消息、搜索聊天记录、管理群聊成员、上传下载图片和文件(支持大文件分片下载)、管理表情回复。当用户需要发消息、查看或搜索聊天记录、下载聊天中的文件、查看群成员、搜索群、创建群聊或话题群、管理标记数据…
v1.0.0 · audit A · 12 stars
lark-mail SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-mail/SKILL.md · 飞书邮箱 — draft, compose, send, reply, forward, read, and search emails; manage drafts, folders, labels, contacts…
v1.0.0 · audit C · 12 stars
lark-markdown SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-markdown/SKILL.md · 飞书 Markdown:查看、创建、上传和编辑 Markdown 文件。当用户需要创建或编辑 Markdown 文件、读取或修改时使用。
v1.0.0 · audit A · 12 stars
lark-minutes SKILL.md skill
kendrick-stein/mcislab_deepread · .agents/skills/lark-minutes/SKILL.md · 飞书妙记:妙记相关基本功能。1.查询妙记列表(按关键词/所有者/参与者/时间范围);2.获取妙记基础信息(标题、封面、时长 等);3.下载妙记音视频文件;4.获取妙记相关 AI…
v1.0.0 · audit A · 12 stars

Every file in kendrick-stein/mcislab_deepread

Other files named memory-distill

memory-distill SKILL.md skill
dannykkh/skill-olympus · skills/memory-distill/SKILL.md · memory/{gotchas,learned}/observations.jsonl raw 관찰을 정제 .md로 변환하는 사용자 트리거 스킬. 핸드오프 의존이 약할 때 의지로 호출. Dreaming-like…
git:20261010.165f741 · audit A · 7 stars

Browse by kind, by grade A, or by owner.