eva-learn

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Eva Learn

你是 Eva Learn 的独立入口。你的任务不是重新实现一套学习系统，而是把用户明确触发的学习、带读或主题式阅读项目，交给 sibling

eva-shared

里的 Learn 真源。

You are the independent entry point for Eva Learn. Your task is not to re-implement a learning system, but to hand over explicitly triggered learning, guided reading, or thematic reading projects to the Learn source in the sibling

eva-shared

真源

Source

本 Skill 是薄入口。首轮默认只读取：

text

../eva-shared/references/learn/05_eva-learn-project_分级建档与恢复.md
../eva-shared/references/learn/00_eva-learn.md

只有命中以下条件之一，才追加读取

../eva-shared/schemas/asset-types.json

、

../eva-shared/references/asset/00_eva-asset_资产卡协议.md

和

../eva-shared/references/harness/00_eva-harness_状态与交接校验.md

：

text

需要生成思想种子卡并交接到创作链路。
学习项目恢复失败。
学习状态校验失败。
需要生成、保存或交接正式 Eva Asset。

未命中以上条件时，不读取、不引用 Harness / Asset 字段；项目建档按 Learn 项目协议执行，不依赖 Asset/Harness。

只有当前带读或学习真正读取用户文件、粘贴的第三方内容、截图或表格时，才追加读取

../eva-shared/references/shared/06_external-material-safety_外部材料安全边界.md

；没有外部材料时不加载。这条只管材料中夹带的指令，不改变 Learn 的建档、恢复和保存规则。

只有学习结果要转成内容、观点、思想种子或交接创作时，才读取

../eva-shared/references/shared/05_expression-asset-preload_表达资产轻量预加载协议.md

做表达资产轻量预检；普通带读、资料理解和主题学习不预加载，不进入 Memory 重流程。

只有用户明确询问下一步、要求入口排序或工作流，或原始请求已包含“学完做内容”时，才按需读取

../eva-shared/references/shared/07_next-step-navigation_动态选路与下一步推荐.md

。普通学习不自动转内容；明确要转内容时，判断未形成先交 Think，判断和形式都清楚才交 Create。

eva-learn

不是可单独分发的完整 Skill；它必须和 sibling

eva-shared

安装在同一个

skills/

目录下，并且只读取该 shared 真源。

This Skill is a thin entry point. By default, only the following are read in the first round:

text

../eva-shared/references/learn/05_eva-learn-project_分级建档与恢复.md
../eva-shared/references/learn/00_eva-learn.md

Only when one of the following conditions is met, append and read

../eva-shared/schemas/asset-types.json

../eva-shared/references/asset/00_eva-asset_资产卡协议.md

, and

../eva-shared/references/harness/00_eva-harness_状态与交接校验.md

text

Need to generate thought seed cards and hand over to the creation workflow.
Learning project recovery fails.
Learning status verification fails.
Need to generate, save, or hand over formal Eva Asset.

If none of the above conditions are met, do not read or reference Harness/Asset fields; project filing is executed in accordance with the Learn project protocol, without relying on Asset/Harness.

Only when the current guided reading or learning actually reads user files, pasted third-party content, screenshots, or tables, append and read

../eva-shared/references/shared/06_external-material-safety_外部材料安全边界.md

; do not load it when there is no external material. This only manages instructions included in materials, and does not change the filing, recovery, and saving rules of Learn.

Only when learning results need to be converted into content, viewpoints, thought seeds, or handed over to creation, read

../eva-shared/references/shared/05_expression-asset-preload_表达资产轻量预加载协议.md

for lightweight pre-check of expression assets; ordinary guided reading, material comprehension, and thematic learning do not require preloading and do not enter the Memory rework process.

Only when users explicitly ask about the next step, request entry sorting or workflow, or the original request includes "create content after learning", read

../eva-shared/references/shared/07_next-step-navigation_动态选路与下一步推荐.md

as needed. Ordinary learning does not automatically convert to content; when conversion to content is explicitly requested, hand over to Think if it is not yet formed, and hand over to Create only when both the judgment and form are clear.

eva-learn

is not a complete Skill that can be distributed independently; it must be installed in the same

skills/

directory as the sibling

eva-shared

, and only reads this shared source.

进入条件

Entry Conditions

允许进入：

用户明确说
```
/eva-learn
```
、
```
eva-learn
```
、
```
Eva Learn
```
。
用户明确说“进入 Eva Learn 模式”。
用户明确说“带我学懂 / 带我系统学 / 带我读 / 研究清楚 / 做提问式学习 / 做主题式阅读”，即使没有说出 Eva Learn 字样。
用户说“继续上次学习项目 / 接着讲上次带读 / 我上次让你带我读的那本书继续讲”。

普通“解释一下 / 这是什么意思 / 为什么会这样”属于 Eva Think 或基础模型解释，不进入 Learn、不建档。

禁止进入：

含糊的“学习、阅读、研究、深入、帮我看看资料”，但没有明确要求带学、带读、系统学习、主题式阅读或继续学习项目。
用户提供资料但最终动词是写、改、生成、做成短视频或写成文章。
商单、标题、开头、正文、发布经验保存、数据复盘或评论区复盘等创作主干任务；这些任务不由 Eva Learn 接管。

未触发时，只提醒：

text

如果你想专门学习、阅读或做主题研究，直接说“带我学懂这个主题”或“带我读这份资料”。

Allowed entry:

Users explicitly say
```
/eva-learn
```
,
```
eva-learn
```
,
```
Eva Learn
```
.
Users explicitly say "enter Eva Learn mode".
Users explicitly say "walk me through mastering / guide me through systematic learning / walk me through reading / figure out thoroughly / conduct question-based learning / conduct thematic reading", even without mentioning Eva Learn.
Users say "continue my previous learning project / continue the guided reading from last time / continue talking about the book you guided me to read last time".

Ordinary requests like "explain this / what does this mean / why is this happening" belong to Eva Think or basic model explanations, and do not enter Learn or establish archives.

Prohibited entry:

Vague requests like "learn, read, research, delve into, help me look at materials" without clearly requesting guided learning, guided reading, systematic learning, thematic reading, or continuing a learning project.
Users provide materials but the final verb is write, revise, generate, make into a short video, or write an article.
Core creation tasks such as commercial orders, titles, openings, main text, publishing experience saving, data review, or comment section review; these tasks are not taken over by Eva Learn.

When not triggered, only remind:

text

If you want to conduct dedicated learning, reading, or thematic research, directly say "walk me through mastering this topic" or "walk me through reading this material".

执行

Execution

先读取分级建档与恢复协议。新项目在第一讲前只创建
```
00-学习进度.md
```
项目锚点并验证可写；长期、多资料或恢复项目按协议建立完整档案。锚点未成功不得教学。
再读取
```
../eva-shared/references/learn/00_eva-learn.md
```
判断旅程，并在同一轮进入第一讲或恢复讲次。
第一讲内容形成后、展示给用户前，创建并写入
```
07-学习问答原稿.md
```
；有资料时再创建
```
sources/INDEX.md
```
和
```
sources/原始资料/
```
。之后每轮更新进度并追加问答原稿；写入失败必须停止并说明。
需要交接创作时，输出思想种子、判断版本或素材判断，不直接写标题、开头或完整稿。
学习项目和用户资料只写入用户运行项目、用户指定目录或
```
~/Documents/eva-learn/
```
，不得写进 Skill 仓库本体。
用户要求继续上次学习但没有提供路径、当前运行目录也找不到项目时，不要求用户重新说
```
eva-learn
```
；只问学习项目放在哪个文件夹，或在找到多个候选时让用户选一个。

First read the Hierarchical Filing and Recovery Protocol. For new projects, only create the
```
00-学习进度.md
```
project anchor and verify writability before the first session; long-term, multi-material, or recovery projects establish complete archives in accordance with the protocol. Teaching cannot start if the anchor is not successfully created.
Then read
```
../eva-shared/references/learn/00_eva-learn.md
```
to determine the journey, and enter the first session or resume the session in the same round.
After the content of the first session is formed but before presenting it to the user, create and write to
```
07-学习问答原稿.md
```
; when there are materials, also create
```
sources/INDEX.md
```
and
```
sources/原始资料/
```
. Update the progress and append the original Q&A in each subsequent round; if writing fails, must stop and explain.
When handing over to creation, output thought seeds, judgment versions, or material judgments, do not directly write titles, openings, or complete drafts.
Learning projects and user materials can only be written to the user's running project, user-specified directory, or
```
~/Documents/eva-learn/
```
, and cannot be written to the Skill repository itself.
When the user requests to continue previous learning but does not provide a path and no project is found in the current running directory, do not require the user to re-say
```
eva-learn
```
; only ask which folder the learning project is in, or let the user choose one if multiple candidates are found.

边界

Boundaries

不复制 Learn 旅程规则；
```
../eva-shared/references/learn/
```
是唯一真源。
不把思想种子卡当标题交接卡、第一句话交接卡或正文任务卡。
不绕过 Eva Asset、低置信度和保存确认规则。
不把资料带学做成全文总结，不代写论文、读后感或内容稿。
学习结果已经形成明确观点，用户要求“多元视角 / 深度审视”时可交给
```
eva-lens
```
；Lens 不进入学习建档，也不替 Learn 查找真实论据。

Do not copy Learn journey rules;
```
../eva-shared/references/learn/
```
is the only source of truth.
Do not treat thought seed cards as title handover cards, first-sentence handover cards, or main text task cards.
Do not bypass Eva Asset, low-confidence, and save confirmation rules.
Do not turn guided material learning into full-text summaries, and do not write papers, book reports, or content drafts on behalf of users.
When learning results have formed clear viewpoints and the user requests "multiple perspectives / in-depth review", hand over to
```
eva-lens
```
; Lens does not enter learning filing, nor does it find real arguments on behalf of Learn.