Token导航 LogoToken导航

Agent Skill · 研究检索

archive-to-brain(存档到大脑)

archive-to-brain 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

archive-to-brain 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为调用外部服务;许可证信息还不明确,第三方扫描有未通过项,按安装量在研究检索类的 73,675 个 Skill 里排第 72,083。

总安装 190周安装 8GitHub Stars 2下载 67核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:archive-to-brain(存档到大脑)
来源仓库:https://github.com/nweii/agent-stuff
仓库路径:skills/archive-to-brain
安装命令:
npx skills add https://github.com/nweii/agent-stuff --skill archive-to-brain
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/nweii/agent-stuff --skill archive-to-brain

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于将 AI 对话摘要保存至 Obsidian 知识库。支持技术工作会话和战略规划等不同类型对话归档。提供思维演进分析和价值判断等专业能力。archive-to-brain 属于研究检索类 Skill,可作为该场景下的辅助能力补充。安装需从指定 GitHub 仓库获取,使用时需确认 Obsidian 路径权限。

涉及知识入库时应注意格式规范和元数据完整性。

来源仓库内路径:skills/archive-to-brain

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/nweii/agent-stuff --skill archive-to-brain;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要调用外部服务,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

archive-to-brain 怎么安装?

在终端执行 npx skills add https://github.com/nweii/agent-stuff --skill archive-to-brain。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

archive-to-brain 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

archive-to-brain 需要什么权限?

权限级别是调用外部服务。该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

archive-to-brain 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

archive-to-brain 用的人多吗?

公开数据是累计安装 190 次,近一周安装 8 次,下载 67 次,GitHub 2 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

archive-to-brain 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
archive-to-brain

存档到大脑 · 本页

1902调用外部服务未明确未通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Archive conversation to Brain vault

Create an archival summary of an AI conversation and save it to Nathan's Obsidian vault (Brain).

Deep analysis requirements

Conduct a thorough analysis of the entire conversation:

  1. Read through completely first, identifying all conceptual threads, task sequences, and transitions
  2. Note patterns in questioning, resistance points, breakthrough moments, or technical hurdles
  3. Identify the conversation's nature (technical work session, creative exploration, strategic planning, philosophical inquiry, etc.)
  4. Understand what made this particular exchange worth preserving (insight-driven vs. action-documentation)
  5. Determine what structure would best capture its unique value (narrative vs. log-formatted)

Look deeply for:

  • The real question beneath the initial question
  • How the problem space was redefined or the technical path was forged
  • Moments where assumptions were challenged or implementation details were decided
  • Conceptual frameworks or technical patterns that emerged organically
  • The emotional/intellectual journey or the step-by-step progress of a work session
  • Valuable tangents or "failed" approaches that taught something or informed the final code
  • Connections made between seemingly unrelated ideas or system components
  • What remained intentionally unresolved or deferred to later tasks

Creating descriptive structure

Instead of using generic headings like "Initial Question" or "Key Findings," create headings that describe the actual content of each section. The heading should give readers immediate context about what happened in that part of the conversation.

Examples of descriptive headings:

  • "Starting from hourly vs. project pricing questions"
  • "Why the recursive function kept hitting memory limits"
  • "Exploring whether this needs to be real-time"
  • "The confusion about state management"
  • "Deciding between complexity and maintainability"

Use sentence-case for headings, not title case. Avoid marketing-speak, dramatic phrasing, or trying to be clever.

Flexible documentation approaches

Let the conversation's natural flow determine your structure:

For problem-solving sessions: Open with what broke/what problem triggered the conversation → Document failed approaches if instructive → Describe the working solution → Note implementation details or next steps

For creative explorations: Start with the initial vision or desire → Show how ideas evolved or branched → Capture key decisions and why they were made → Preserve unexplored directions worth revisiting

For learning journeys: Begin with what the user didn't understand → Track how understanding built piece by piece → Highlight breakthrough moments → List remaining questions

For work sessions & implementation logs: Define the session's objective → Document specific actions taken and files modified → Capture technical hurdles and how they were resolved → Summarize the current state of the work and remaining tasks

For strategic thinking: Frame the decision that needed making → Explore options considered and their trade-offs → Document the framework or criteria that emerged → Capture action items or next considerations

Excerpt guidelines

Include conversation excerpts that show thinking in action:

Nathan: "[moment of recognition or confusion]" AI: "[response that shifted understanding or articulated key insight]"

Choose excerpts that reveal intellectual movement — the moments where thinking actually changed, not just where information was exchanged. Be generous in your excerpt lengths.

Naming convention

  • Format: {{Type}} - {{topic}} YYYY-MM.md
  • Use Thinking for insight-heavy, reflective, or exploratory conversations
  • Use Log for action-leaning work sessions, implementation logs, or technical sessions
  • Example: Thinking - Portfolio strategy 2025-08.md
  • Example: Log - Refactoring auth middleware 2025-01.md

Vault save logic

1. Compose the frontmatter

Use today's date for created, modified, and last.

---
aliases:
  - [1-2 intuitive alternative titles in sentence case, e.g. "Thinking through X" or "Notes from Y session"]
categories: "[[Thinking]]"
type:
icon: [camelCase Lucide icon name prefixed with "Li" that fits the topic, e.g. LiBrainCircuit, LiMessageCircle, LiCode2]
publish: false
description: [1–2 sentence summary of what this conversation covered and why it was worth saving]
last: YYYY-MM-DD
tags:
  - thinking
  - [2-3 additional tags reflecting the specific topic, domain, or people involved]
related:
  - ["[[Note name]]"] # only include confirmed vault notes that came up in conversation; wikilinks require quotes in YAML arrays
# Optional: prev / next wikilinks for continuity with adjacent Thinking or Log notes — see paragraph after this block
created: YYYY-MM-DD
modified: YYYY-MM-DDTHH:mm
---

For Log notes, change the tags entry to logs and update categories to "[[Logs]]" if appropriate. Generate aliases, description, icon, and tags from the actual conversation content — they should aid recall beyond the filename. Only populate related with notes you have confirmed exist in the vault; leave the array empty if none were referenced. Do not put a note in related if it is already tied via prev or next — same link twice is redundant; use related for broader or non-adjacent ties.

Continuity (prev / next): When the conversation surfaces another Thinking or Log note as the immediate predecessor or successor, or the user names one, treat that as a continuity link — not only related. Set prev and/or next on the new note per the vault’s AGENTS.md, and update the other note(s) the same way so the chain stays bidirectional (e.g. if this archive follows note A, set this note’s prev to A and set A’s next to this note). Use whatever the environment supports: Obsidian CLI property:set, editing YAML frontmatter in place (after create), notesmd-cli if it exposes properties, vault MCP, etc. If the new note sits between two existing notes, fix all three. Omit prev / next when no clear adjacent note exists. Typical case for a new archive: this note is usually the next after the prior session’s Thinking/Log; set prev on the new file and patch the previous note’s next.

2. Determine the save folder

  • Personal life, emotions, identity, relationships, dreams, health → 03-Records/Journaling
  • Work, projects, productivity (including personal productivity), technical sessions, client work → 03-Records/Working

Within 03-Records/Working, check for existing subfolders that match the conversation topic (search the vault, list the directory on disk, or use whatever tooling the host provides). Use a matching subfolder when it clearly fits; otherwise save to the root of 03-Records/Working. Don't assume which subfolders exist — they change over time.

If a specific save location was provided, use it directly.

3. Save to vault

Pick a write path that matches the machine: desktop with Obsidian open, headless server with a synced vault folder, or chat-only with no filesystem. No single tool is always present.

Preferred order when multiple options exist (skip steps that are unavailable):

  1. Obsidian CLI — Best when the Obsidian app can run (indexing, property:set, wikilink-style file=). Use vault=<name> if there are multiple vaults. Example create: obsidian vault=Brain create name="{{filename}}" path="{{folder}}/{{filename}}.md" content="{{full content with frontmatter}}" Example continuity (bidirectional prev / next): obsidian vault=Brain property:set file="{{filename}}" name=prev value='[[Prior note title]]' obsidian vault=Brain property:set file="Prior note title" name=next value='[[{{filename}}]]'
  2. notesmd-cli — Typical on headless hosts where the vault is synced but Obsidian's CLI is not available. Use its create/update commands per that tool’s --help; set prev / next by editing frontmatter if the CLI does not expose properties.
  3. Direct read/write — If the absolute vault root is known and writable, write {{folder}}/{{filename}}.md with the full file contents (including frontmatter), then patch the prior note’s frontmatter for next if needed.
  4. Vault MCP — Use when the agent only has MCP filesystem access to the vault and no local CLI. Prefer create/move/update there over leaving the archive only in chat.
  5. Manual handoff — Output the full note in a markdown code block with the intended path above it so the user can paste or save it.

Pipe multiline content via stdin or use \n escaping in CLI content= values when required.

4. Log to the corresponding daily note

After saving the archive, append a bullet to the ## Log section of the daily note that best corresponds to when the conversation actually took place — usually today, but not always. Use your understanding of the conversation's content and timing as the primary guide; the archive's date or last frontmatter properties are useful supporting signals if the date is ambiguous.

Format:

- Archived [[Note filename without extension]] — [one sentence: what the conversation covered and why it was saved]

The summary should be a tight, plain-language sentence — no marketing language. Pull it from the note's description frontmatter or compose it from the archive's subject matter.

Resolving the target daily note — derive the path from the chosen date:

  • Daily notes follow 01-Days/YYYY-MM-DD-ShortDayName.md (e.g. 01-Days/2025-08-14-Thu.md)
  • Derive the short day name from the date (Mon, Tue, Wed, Thu, Fri, Sat, Sun)

How to update it (use whichever is available):

  • Obsidian CLI: Read the target note by path or title, locate ## Log, and write the updated file with the bullet appended.
  • Direct file edit: Read 01-Days/YYYY-MM-DD-ShortDayName.md, find ## Log, and append the bullet on a new line after the last existing bullet in that section.
  • Vault MCP: Use update/edit tools to insert the bullet into ## Log.

If the target daily note does not exist, skip this step rather than creating it.

Remember

  • You're documenting intellectual exploration OR technical execution/work sessions
  • Perform deep analysis to identify all important threads, transitions, and task sequences
  • Use headings that describe what actually happened or what was achieved in that section
  • Keep language natural and straightforward — no marketing-speak or forced drama
  • Capture why this journey or work session matters, and what was actually produced or decided
  • Include the messy, human elements — confusion, recognition, technical frustrations, breakthroughs
  • Preserve what would be valuable to revisit months or years later
  • Use third person or neutral documentation style, not first person (except when quoting)

平台分布

Codex

38.13%
按下载量换算26

Claude

28.59%
按下载量换算19

Cursor

19.92%
按下载量换算13

Gemini CLI

9.52%
按下载量换算6

安全审计

Gen Agent Trust Hub

可疑

Socket

通过

Snyk

未通过

权限和风险

调用外部服务

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills