Content Digest
μ½ν
μΈ β Quiz-First νμ΅ β μ νμ κΉμ΄ νμ β κ·Όλ³Έ κ°λ
νμ₯.
Task Agent κΈ°λ° μ€κ³ : κΈ΄ 컨ν
μ€νΈλ subagentκ° μ²λ¦¬νκ³ , λ©μΈ μΈμ
μ μ΅μ’
κ²°λ‘ λ§ μλΉ
μν€ν
μ² μμΉ
Context Separation : κΈ΄ μλ§/λ³Έλ¬Έμ Task agentκ° μ²λ¦¬, λ©μΈ μΈμ
μ κ°λ²Όμ΄ md νμΌλ§ Read
Clean Transcript : μλ§μμ λ²νΈ, μκ° μ κ±° β μμ μμ΄ ν
μ€νΈλ§ μΆμΆ
Web Research Integration : μΆμΆλ ν€μλλ‘ μλ μΉ λ¦¬μμΉ
Single Output : λͺ¨λ μ²λ¦¬ κ²°κ³Όλ λ¨μΌ md νμΌλ‘ μ μ₯
μ§μ μ½ν
μΈ
νμ
μΆμΆ λ°©λ²
μ μ₯ κ²½λ‘
YouTube
Task agent (yt-dlp + μ μ )
research/digests/youtube/
X/Twitter
fetch-tweet μ€ν¬ (api.fxtwitter.com)
research/digests/tweet/
Webpage
Task agent (browser + μ μ )
research/digests/web/
PDF
Task agent (Read + μ μ )
research/digests/pdf/
ν΅μ¬ μμΉ
Quiz-First : μμ½ λ³΄κΈ° μ μ ν΄μ¦λΆν° (Pretesting Effect β 9-12% ν₯μ)
Knowledge Gap : νλ¦° λ¬Έμ κ° νΈκΈ°μ¬μ λ§λ€κ³ , νΈκΈ°μ¬μ΄ κΈ°μ΅μ κ°ν
μ νμ κΉμ΄ : μ¬μ©μκ° λ μκ³ μΆμ λΆλΆλ§ κΉκ²
κ·Όλ³Έ νμ₯ : μ½ν
μΈ λλ¨Έμ κΈ°μ΄ κ°λ
κΉμ§ μΉ κ²μμΌλ‘ νμ₯
μν¬νλ‘μ° κ°μ (Task Agent κΈ°λ°)
Phase 1: μ½ν
μΈ νμ
κ°μ§
Phase 2: Task Agent μ€ν (μ½ν
μΈ μΆμΆ + μ μ + μΉ λ¦¬μμΉ + md μ μ₯)
Phase 3: λ©μΈ μΈμ
μμ κ²°κ³Ό md Read
Phase 4: Pre-Quiz (3λ¬Έμ )
Phase 5: μ νμ μ½ν
μΈ μ 곡
Phase 6: λ³Έ ν΄μ¦ (9λ¬Έμ )
Phase 7: Elaborative Interrogation
Phase 8: Foundation Expansion
Phase 9: μ€ν€λ§ μ°κ²°
Phase 10: λ¬Έμ μ
λ°μ΄νΈ (ν΄μ¦ κ²°κ³Ό λ°μ)
Phase 11: νμ μ ν
Phase 1: μ½ν
μΈ νμ
κ°μ§
μ
λ ₯ ν¨ν΄μ λ°λΌ μ½ν
μΈ νμ
μλ κ²°μ :
ν¨ν΄
νμ
youtube.com, youtu.be
YouTube
x.com, twitter.com
X/Twitter
http://, https:// (κΈ°ν)
Webpage
.pdf νμΌ κ²½λ‘
PDF
λͺ
ννμ§ μμΌλ©΄ μ¬μ©μμκ² νμΈ:
AskUserQuestion:
questions:
- question: "μ΄λ€ μ½ν
μΈ λ₯Ό λΆμν κΉμ?"
header: "Type"
options:
- label: "YouTube μμ"
description: "URLμ μλ €μ£ΌμΈμ"
- label: "μΉνμ΄μ§/μν°ν΄"
description: "URLμ μλ €μ£ΌμΈμ"
- label: "PDF λ¬Έμ"
description: "νμΌ κ²½λ‘λ₯Ό μλ €μ£ΌμΈμ"
Phase 2: Task Agent μ€ν (ν΅μ¬)
λ©μΈ μΈμ
μ contextλ₯Ό 보νΈνλ©΄μ κΈ΄ μ½ν
μΈ λ₯Ό μ²λ¦¬
2-1. Task Agent νΈμΆ ν¨ν΄
Task:
subagent_type: "general-purpose"
description: "μ½ν
μΈ μΆμΆ λ° λΆμ"
prompt: |
## λͺ©ν
{URL/νμΌκ²½λ‘}μμ μ½ν
μΈ λ₯Ό μΆμΆνκ³ λΆμνμ¬ md νμΌλ‘ μ μ₯
## λ¨κ³ (μμ μ€μ)
1. μ½ν
μΈ μΆμΆ (νμ
λ³ λ°©λ² μ μ©)
2. ν
μ€νΈ μ μ (λ²νΈ, μκ° μ κ±° β μμ΄λ§ μΆμΆ)
3. ν΅μ¬ ν€μλ μΆμΆ (5-10κ°)
4. μΉ λ¦¬μμΉ (ν€μλλ³ WebSearch)
5. **ν΅μ¬ μμ½ μμ±** (3-5λ¬Έμ₯)
6. **μ£Όμ μΈμ¬μ΄νΈ λμΆ** (3κ°)
7. **ν΄μ¦ μ¬λ£ μμ±** (μμ½/μΈμ¬μ΄νΈ κΈ°λ°μΌλ‘ ν΅μ¬ μ£Όμ λ§)
8. md νμΌ μ μ₯
## μΆλ ₯ κ²½λ‘
research/digests/{type}/{YYYY-MM-DD}-{sanitized-title}.md
2-2. X/Twitter μΆμΆ (fetch-tweet μ€ν¬ νμ©)
Task Agent λΆνμ - fetch-tweet μ€ν¬λ¦½νΈλ‘ μ§μ μΆμΆ (μ§§μ μ½ν
μΈ )
python3 .claude/skills/fetch-tweet/scripts/fetch_tweet.py "{URL}" --json
JSON μλ΅μμ νμ©ν νλ:
tweet.text: νΈμ λ³Έλ¬Έ
tweet.author: μμ±μ μ 보 (name, bio, followers)
tweet.likes/retweets/views: μΈκ²μ΄μ§λ¨ΌνΈ
tweet.quote: μΈμ© νΈμ (μμ κ²½μ° λμΌ κ΅¬μ‘°)
tweet.media: μ²¨λΆ μ΄λ―Έμ§/μμ
νΈμμ μ§§μΌλ―λ‘ Task Agent μμ΄ λ©μΈ μΈμ
μμ μ§μ μ²λ¦¬.
μΈμ© νΈμμ΄ μμΌλ©΄ ν¨κ» ν¬ν¨νμ¬ λΆμ.
μ μ₯ κ²½λ‘: research/digests/tweet/{YYYY-MM-DD}-{author}-{short-topic}.md
2-3. YouTube μΆμΆ (Task Agent λ΄λΆ)
yt-dlp --write-auto-sub --sub-lang "en" --skip-download \
--convert-subs vtt -o "%(title)s" "{URL}"
sed -E 's/^[0-9]+$//' | \
sed -E 's/[0-9]{2}:[0-9]{2}:[0-9]{2}.*//g' | \
sed -E 's/<[^>]+>//g' | \
tr -s '\n' | \
grep -v '^$'
μ μ κ²°κ³Ό: μμ μμ΄ ν
μ€νΈλ§ λ¨μ (μκ°, λ²νΈ, μ€λ³΅ μμ)
2-4. Webpage μΆμΆ (Task Agent λ΄λΆ)
1. mcp__claude-in-chrome__tabs_context_mcp
2. mcp__claude-in-chrome__tabs_create_mcp
3. mcp__claude-in-chrome__navigate: url="{URL}"
4. mcp__claude-in-chrome__get_page_text: tabId={tabId}
5. μ€ν¬λ‘€ ν μΆκ° μ½ν
μΈ νμΈ
2-5. PDF μΆμΆ (Task Agent λ΄λΆ)
Read: file_path="{PDF κ²½λ‘}"
2-6. μΉ λ¦¬μμΉ (Task Agent λ΄λΆ)
μΆμΆλ ν
μ€νΈμμ ν΅μ¬ ν€μλ 5-10κ° μλ³ ν:
WebSearch (λ³λ ¬ μ€ν):
- "{ν€μλ1} explained"
- "{ν€μλ2} research"
- "{μ μ/λ°νμ} {μ£Όμ }"
- "{ν΅μ¬κ°λ
} fundamentals"
2-7. μ΅μ’
md νμΌ μ μ₯ (Task Agent λ΄λΆ)
κ²½λ‘: research/digests/{type}/{YYYY-MM-DD}-{sanitized-title}.md
---
title : { μ½ν
μΈ μ λͺ©}
type : { youtube| web| pdf}
url : { URL λλ νμΌκ²½λ‘}
author : { μ μ/μ±λλͺ
}
date : { λ°ν λ μ§}
processed_at : { μ²λ¦¬ μΌμ}
keywords : [ { ν€μλ1} , { ν€μλ2} , ... ]
---
# {μ½ν
μΈ μ λͺ©}
## ν΅μ¬ μμ½
{3-5λ¬Έμ₯ μμ½}
## μ£Όμ μΈμ¬μ΄νΈ
1. ** {μΈμ¬μ΄νΈ1} ** : μ€λͺ
2. ** {μΈμ¬μ΄νΈ2} ** : μ€λͺ
3. ** {μΈμ¬μ΄νΈ3} ** : μ€λͺ
## μΉ λ¦¬μμΉ κ²°κ³Ό
### {ν€μλ1}
- λ°κ²¬ λ΄μ© μμ½
- μΆμ²: {URL}
### {ν€μλ2}
- λ°κ²¬ λ΄μ© μμ½
- μΆμ²: {URL}
## μλ¬Έ (μ μ λ¨)
{λ²νΈ/μκ° μ κ±°λ μμ ν
μ€νΈ}
## Quiz μ¬λ£ (Pre-Quiz + λ³Έ Quizμ©)
> ** μμ± μμ ** : λ°λμ μμ "ν΅μ¬ μμ½"κ³Ό "μ£Όμ μΈμ¬μ΄νΈ"λ₯Ό λ¨Όμ μμ±ν ν, μ΄λ₯Ό κΈ°λ°μΌλ‘ ν΄μ¦ μμ±
> ** μΆμ μμΉ ** : ν΅μ¬ μ£Όμ λ§ μΆμ . λ μ§, ν΅κ³, μ§μ½μ μΈλΆμ¬ν μ μΈ.
### κΈ°λ³Έ λ 벨 (3λ¬Έμ ν보)
- Q1: {ν΅μ¬ κ°λ
/λ©μμ§ κ΄λ ¨}
- Q2: {μ£Όμ μμΉ κ΄λ ¨}
- Q3: {μ μ ν΅μ¬ μ£Όμ₯ κ΄λ ¨}
### μ€κΈ λ 벨 (3λ¬Έμ ν보)
- Q4: {κ°λ
κ° κ΄κ³}
- Q5: {κ·Όκ±°μ λ
Όλ¦¬ μ°κ²°}
- Q6: {ν΅μ¬ μμ΄λμ΄ λΉκ΅}
### μ¬ν λ 벨 (3λ¬Έμ ν보)
- Q7: {μ€μ μ μ©/μμ©}
- Q8: {ν΅μ¬ μ리μ νμ₯}
- Q9: {μ μ κ΄μ μ ν¨μ}
Phase 3: λ©μΈ μΈμ
μμ κ²°κ³Ό Read
Task Agent μλ£ ν:
Read: file_path="research/digests/{type}/{YYYY-MM-DD}-{sanitized-title}.md"
λ©μΈ μΈμ
μ μ μ λ md νμΌλ§ μ½μ β context ν¨μ¨ κ·Ήλν
Phase 4: Pre-Quiz (ν΅μ¬)
λͺ©μ : μ 보 κ° μμ± β μ£Όμλ ₯ νλΌμ΄λ° β λ₯λμ νμ΅ μ λ
ν΄μ¦ μΆμ μμΉ
ν΅μ¬ μ£Όμ λ§ μ§λ¬Έ : μ¬μν μΈλΆμ¬νμ΄λ μ«μκ° μλ, μ½ν
μΈ μ ν΅μ¬ λ©μμ§μ μ§κ²°λλ λ΄μ©λ§ μΆμ
β
ν΅μ¬ κ°λ
, μ£Όμ μμΉ, μ μμ ν΅μ¬ μ£Όμ₯
β λ μ§, ν΅κ³ μμΉ, λΆμμ μμ, μ§μ½μ μΈλΆμ¬ν
κ²°κ³Ό md νμΌμ "Quiz μ¬λ£" μΉμ
μ νμ©νμ¬ 3λ¬Έμ μΆμ :
AskUserQuestion:
questions:
- question: "[Pre-Quiz] μ΄ μ½ν
μΈ μμ λ€λ£° κ² κ°μ ν΅μ¬ κ°λ
μ?"
header: "PQ1"
options: [4κ° μ νμ§]
- question: "[Pre-Quiz] μ μκ° κ°μ‘°ν κ² κ°μ λ©μμ§λ?"
header: "PQ2"
options: [4κ° μ νμ§]
- question: "[Pre-Quiz] μ΄ μ£Όμ μμ κ°μ₯ μ€μν μμΉμ?"
header: "PQ3"
options: [4κ° μ νμ§]
κ²°κ³Ό μ²λ¦¬ :
μ λ΅/μ€λ΅ μ¦μ νμ
νλ¦° λ¬Έμ β "μ΄ λΆλΆμ μ½ν
μΈ μμ νμΈν΄λ³΄μΈμ" μλ΄
Knowledge Gap μμ± : "μ΄μ μ½ν
μΈ λ₯Ό 보면 λ΅μ μ°Ύκ³ μΆμ΄μ§ κ²μ
λλ€"
Phase 5: μ νμ μ½ν
μΈ μ 곡
Pre-Quiz κ²°κ³Όμ λ°λΌ μ¬μ©μμκ² μ νμ§ μ 곡:
AskUserQuestion:
questions:
- question: "μ΄λ€ μ½ν
μΈ λ₯Ό λ¨Όμ 보μκ² μ΅λκΉ?"
header: "Content"
options:
- label: "νλ¦° λ¬Έμ κ΄λ ¨ μΉμ
λ§"
description: "Pre-Quizμμ νλ¦° λΆλΆμ λ΅μ μ°Ύμ보기"
- label: "ν΅μ¬ μΈμ¬μ΄νΈ 3κ°"
description: "μ½ν
μΈ μ κ°μ₯ μ€μν ν¬μΈνΈλ§"
- label: "μ 체 μμ½ + μΈμ¬μ΄νΈ"
description: "μ’
ν©μ μΈ μ½ν
μΈ λΆμ"
- label: "λ°λ‘ λ³Έ ν΄μ¦λ‘"
description: "μμ½ μμ΄ 9λ¬Έμ ν΄μ¦ μ§ν"
5-1. νλ¦° λ¬Έμ κ΄λ ¨ μΉμ
Pre-Quiz μ€λ΅κ³Ό κ΄λ ¨λ μΉμ
λ§ μΆμΆ:
YouTube: ν΄λΉ νμμ€ν¬ν
Webpage: κ΄λ ¨ λ¨λ½
PDF: ν΄λΉ νμ΄μ§/μΉμ
5-2. ν΅μ¬ μΈμ¬μ΄νΈ (κ°κ²° λͺ¨λ)
## ν΅μ¬ μΈμ¬μ΄νΈ 3κ°
1. ** [ν€μλ] ** : 1-2λ¬Έμ₯ μ€λͺ
2. ** [ν€μλ] ** : 1-2λ¬Έμ₯ μ€λͺ
3. ** [ν€μλ] ** : 1-2λ¬Έμ₯ μ€λͺ
5-3. μ 체 μμ½ + μΈμ¬μ΄νΈ
## μμ½
{3-5λ¬Έμ₯}
## μΈμ¬μ΄νΈ
### ν΅μ¬ μμ΄λμ΄
### μ μ© κ°λ₯ν μ
Phase 6: λ³Έ ν΄μ¦ (9λ¬Έμ )
3λ¨κ³ Γ 3λ¬Έμ . AskUserQuestionμΌλ‘ κ° λ¨κ³ μ§ν.
μΆμ μμΉ : λͺ¨λ λ¬Έμ λ μ½ν
μΈ μ ν΅μ¬ μ£Όμ μ μ§κ²°λμ΄μΌ ν¨. μ§μ½μ μΈλΆμ¬ν, λ μ§, ν΅κ³ μμΉλ μΆμ κΈμ§.
λ¨κ³
λμ΄λ
μΆμ κΈ°μ€
1
κΈ°λ³Έ
ν΅μ¬ λ©μμ§, μ£Όμ κ°λ
2
μ€κΈ
κ°λ
κ° κ΄κ³, κ·Όκ±° μ°κ²°
3
μ¬ν
μ¬λ‘ λΆμ, μ μ©, ꡬ체μ λ°μ΄ν°
λ¬Έμ μ ν μμΈ: references/quiz-patterns.md
μ¦κ° νΌλλ°± : κ° λ¨κ³ μλ£ ν μ λ΅/ν΄μ€ μ¦μ μ 곡
Phase 7: Elaborative Interrogation
"μ?" μ§λ¬Έμ΄ κΉμ μ²λ¦¬λ₯Ό μ λ° (76% vs 69% μ λ΅λ₯ ν₯μ)
ν΄μ¦ μλ£ ν, ν΅μ¬ κ°λ
μ λν΄ μ¬ν μ§λ¬Έ:
AskUserQuestion:
questions:
- question: "λ€μ μ€ λ κΉμ΄ μ΄ν΄νκ³ μΆμ κ°λ
μ?"
header: "Deep Dive"
multiSelect: true
options:
- label: "{κ°λ
A}"
description: "μ μ΄κ²μ΄ μ€μνμ§ νꡬ"
- label: "{κ°λ
B}"
description: "μ΄κ²μ κ·Όλ³Έ μ리 μ΄ν΄"
- label: "{κ°λ
C}"
description: "μ€μ μ μ© μ¬λ‘ νμ₯"
- label: "λ°λ‘ λ€μ λ¨κ³λ‘"
description: "νμ¬ μ΄ν΄ μμ€μΌλ‘ μΆ©λΆ"
μ νλ κ°λ
μ λν΄:
"μ μ΄κ²μ΄ μ¬μ€μΈκ°?" μ§λ¬Έκ³Ό λ΅λ³
μ½ν
μΈ λ΄ κ·Όκ±° μμΉ (νμμ€ν¬ν/νμ΄μ§/μΉμ
)
μΉ κ²μμΌλ‘ μΆκ° λ§₯λ½ μ 곡
Phase 8: Foundation Expansion (κ·Όλ³Έ νμ₯)
μ½ν
μΈ λλ¨Έμ κΈ°μ΄ μ§μ νμ₯
8-1. κΈ°μ΄ κ°λ
μΉ κ²μ (WebSearch λ³λ ¬ 3-5κ°)
κ²μ 쿼리:
- "{ν΅μ¬ κ°λ
} fundamentals explained"
- "{ν΅μ¬ κ°λ
} κΈ°μ΄ μ리"
- "{μ΄λ‘ /λ°©λ²λ‘ } research paper original"
- "{μ μ/λ°νμ} other works recommendations"
8-2. κ·Όλ³Έ μ§μ μ 리
## Foundation Expansion
### μ΄ μ½ν
μΈ μ κΈ°μ΄κ° λλ κ°λ
λ€
| ------ | ------ | Implementation Guide Prerequisites
βΊ Claude Desktop or compatible AI client with skill support βΊ Clear understanding of task or problem to solve βΊ Willingness to iterate and refine outputs Time Estimate
15-45 minutes depending on use case complexity
Steps
1 Install skill using provided installation command 2 Test with simple use case relevant to your work 3 Evaluate output quality and relevance 4 Iterate on prompts to improve results 5 Integrate into regular workflow if valuable Common Pitfalls
β Expecting perfect results without iteration β Not providing enough context in prompts β Using skill for tasks outside its intended scope β Accepting outputs without review and validation Best Practices β Do
+ Start with clear, specific prompts + Provide relevant context and constraints + Review and refine all outputs before using + Iterate to improve output quality + Document successful prompt patterns β Don't
β Don't use without understanding skill limitations β Don't skip validation of outputs β Don't share sensitive information in prompts β Don't expect skill to replace human judgment π‘ Pro Tips
β
Be specific about desired format and style β
Ask for multiple options to choose from β
Request explanations to understand reasoning β
Combine AI efficiency with human expertise When to Use This β Use when
Use when skill capabilities match your task, clear ROI on time saved, and you can validate outputs. Best for repetitive tasks, learning, and quality improvement.
β Avoid when
Avoid when task requires deep expertise you can't validate, involves sensitive decisions, or when learning process is more valuable than speed of completion.
Learning Path 1 Familiarize yourself with skill capabilities and limitations 2 Start with low-risk, non-critical tasks 3 Progress to more complex and valuable use cases 4 Build expertise through regular use and experimentation Reviews 4.6 β
β
β
β
β
59 reviews
A
Amina Zhang β
β
β
β
β
Dec 28, 2024
Useful defaults in content-digest β fewer surprises than typical one-off scripts, and it plays nicely with `npx skills` flows.
T
Tariq Park β
β
β
β
β
Dec 28, 2024
content-digest fits our agent workflows well β practical, well scoped, and easy to wire into existing repos.
C
Chen Zhang β
β
β
β
β
Dec 24, 2024
I recommend content-digest for anyone iterating fast on agent tooling; clear intent and a small, reviewable surface area.
P
Pratham Ware β
β
β
β
β
Dec 20, 2024
Registry listing for content-digest matched our evaluation β installs cleanly and behaves as described in the markdown.
D
Dhruvi Jain β
β
β
β
β
Dec 16, 2024
We added content-digest from the explainx registry; install was straightforward and the SKILL.md answered most questions upfront.
T
Tariq Haddad β
β
β
β
β
Dec 16, 2024
Solid pick for teams standardizing on skills: content-digest is focused, and the summary matches what you get after install.
K
Kofi Jain β
β
β
β
β
Nov 19, 2024
I recommend content-digest for anyone iterating fast on agent tooling; clear intent and a small, reviewable surface area.
A
Amina Brown β
β
β
β
β
Nov 19, 2024
We added content-digest from the explainx registry; install was straightforward and the SKILL.md answered most questions upfront.
J
James Harris β
β
β
β
β
Nov 15, 2024
Useful defaults in content-digest β fewer surprises than typical one-off scripts, and it plays nicely with `npx skills` flows.
O
Oshnikdeep β
β
β
β
β
Nov 7, 2024
content-digest fits our agent workflows well β practical, well scoped, and easy to wire into existing repos.
showing 1-10 of 59
prev 1 / 6 next
Discussion Comments β not star reviews No comments yet β start the thread.