clawlite-video-content-engine

中文：将 YouTube 视频转化为 ClawLite 营销资产的内容引擎，覆盖摘要、短视频脚本、X thread 与博客输出，支持知识型内容复用与渠道化分发。 日本語：YouTube動画をClawLite向け教育・マーケ配信資産へ再構成。要約、短尺動画、Xスレッド、ブログ出力まで一貫して最適化。 한국어：You...

Security Scan

The skill's instructions reasonably match a YouTube→marketing workflow, but they reference browser UI automation and tools (NotebookLM automation, yt-dlp, screenshot workflows, writing stable folders) without declaring required binaries, runtime capabilities, or credentials — that mismatch warrants caution.

Purpose & Capability

The skill is described as a YouTube→marketing content engine, which fits the included templates and output expectations. However, the SKILL.md explicitly recommends using yt-dlp for raw extraction and includes detailed NotebookLM UI automation JavaScript snippets. The metadata declares no required binaries, no install steps, and no environment variables. That is an incoherence: DOM/UI automation and yt-dlp imply the agent needs browser-automation capabilities, a headless browser or user browser session, and/or a local yt-dlp binary — none of which are declared. Asking to write to a stable output folder is also a capability requirement not documented in the skill metadata.

Instruction Scope

The instructions include precise UI automation guidance (prototype setter for textareas, querySelector examples, click rules, modal detection) and a screenshot-first workflow. Those DOM-manipulation snippets are specific to NotebookLM but could be reused to automate other web UIs if executed by an agent with web-control capabilities. The skill also instructs writing files to a stable folder structure (metadata.json, multiple md files). The SKILL.md does not limit or clarify what runtime will perform these actions (human-in-the-loop vs autonomous browser automation), nor does it require or request NotebookLM credentials. This gives the agent broad discretion to access web UIs and file system state that isn't documented.

Install Mechanism

There is no install spec and no code files beyond documentation, which minimizes the risk of arbitrary installs. That said, the workflow expects external tooling (yt-dlp, a browser automation capability or headless browser) but doesn't provide an install path or declare them as required. The absence of an install spec is safe by itself, but it increases incoherence because the instructions assume tooling the metadata doesn't declare.

Credentials

The skill requests no environment variables or credentials, which is good from a secrets-exfiltration standpoint. However, it assumes access to NotebookLM (a web product) and the user's ability to run UI automation or to be logged into NotebookLM in a browser session. Those implicit dependencies (authenticated browser session, access to YouTube, local yt-dlp binary) are not represented in requires.env or required binaries and are therefore not explicitly justified by the metadata.

Persistence & Privilege

The skill does not request persistent inclusion (always:false) and does not attempt to modify other skills or global agent settings in the documentation. It instructs creating stable output folders, which is a normal operational detail and not itself privileged. Autonomous invocation is allowed by default, but there is no additional privilege escalation or always:true setting.

What to consider before installing

This skill appears to be a well-documented content-pack workflow, but it relies on implicit runtime capabilities that are not declared. Before installing or running it, ask the provider or your platform operator: (1) Will the agent run UI/browser automation (and if so, which tool — Puppeteer, Playwright, a real browser with a user session)? (2) Is yt-dlp or a transcript-extraction tool available on the runtime, or do you need to install one? (3) Where will the generated files be written and who can access that folder? (4) Will NotebookLM be invoked via an authenticated user session (risk: screenshots and injected text can expose other user data)? If you want to proceed, require the skill to declare its runtime dependencies (binaries, required capabilities, and whether it needs interactive browser access) and limit its web-automation scope (explicit target domains only). If you cannot confirm those details, treat the skill as potentially risky to grant web automation or file-write privileges to.

Use this skill to convert third-party educational videos into ClawLite-compatible educational marketing content.

Core principle:

• do not treat source videos as raw material for plagiarism or blind reposting
• treat source videos as learning inputs that become:
  • beginner summaries
  • practical takeaways
  • explainer shorts
  • X threads
  • LinkedIn/Facebook posts
  • short blog summaries
  • soft ClawLite bridge content

Outcome

Turn one source video into a content pack:

• 1 source summary
• 1 beginner translation
• 1 short-form video script
• 1 X thread
• 1 LinkedIn/Facebook post
• 1 short blog summary
• 1 CTA bridge to ClawLite

Output location rule

Write outputs to a stable folder so the workflow is reusable and auditable.

Recommended structure:

video-content/
  <videoId>/
    raw-transcript.md
    notebooklm-summary.md
    jk-marketing-asset.md
    source-note.md
    short-video-script.md
    x-thread.md
    linkedin-post.md
    blog-summary.md
    metadata.json

At minimum, write:

• notebooklm-summary.md
• jk-marketing-asset.md
• source-note.md
• short-video-script.md
• x-thread.md
• blog-summary.md
• metadata.json

Workflow

Normalization rule

NotebookLM output is not the final downstream input. It must be normalized into a JK / marketing-assets layer before Elon, Tony, or Jenny consume it.

Use this chain:

• YouTube / transcript source
• raw extraction layer (for example yt-dlp)
• NotebookLM understanding layer
• JK marketing asset layer
• Elon / Tony / Jenny execution outputs

1. Capture the source video context

Record:

• title
• creator
• URL
• publish date if useful
• duration
• main topic
• likely beginner pain point

If NotebookLM is available, use it for transcript + summary extraction. If NotebookLM is unavailable, create the structure manually from transcript/notes.

When using NotebookLM UI automation:

• use a screenshot-first workflow
• verify the exact input field before typing
• avoid generic textarea selectors
• confirm source creation before moving to content generation

Read references/notebooklm-automation-guide.md before automating NotebookLM.

2. Build a source note

Create a structured source note with:

• what the video is about
• 3 key takeaways
• strongest quote or idea
• why it matters for beginners
• where setup friction appears
• how ClawLite naturally bridges the gap

Read references/source-note-template.md when building the note.

3. Normalize into JK marketing assets

Convert the source + NotebookLM understanding into a reusable asset note for downstream lanes.

The JK asset should include:

• source context
• pain point
• beginner misunderstanding
• 3 key takeaways
• strongest idea / quote
• angle candidates
• hook candidates
• ClawLite bridge
• Elon social angle
• Tony blog angle
• Jenny lifecycle angle
• source / proof lines

This asset layer should become the shared substrate for downstream content generation.

Read references/jk-marketing-asset-template.md when building this layer.

4. Translate the source into ClawLite angles

Do not simply restate the creator video. Create one or more of these angles:

• beginner translation
• practical summary
• “what matters most” summary
• “3 takeaways” summary
• “too long, didn’t watch” summary
• setup-friction reframing

Read references/angle-framework.md when choosing the angle.

5. Create the short-video script

Write a 30–90 second short video script with:

• hook
• 2–3 insights
• beginner framing
• soft ClawLite bridge
• CTA

Prefer:

• educational tone
• real user pain
• concise and clear subtitles
• no hard sell in the first half

Read references/short-video-template.md when writing the script.

6. Expand into a multi-channel content pack

Derive from the same source note and JK marketing asset:

• X thread
• LinkedIn/Facebook post
• short blog summary
• optional newsletter blurb

Read references/content-pack-template.md for the output structure.

7. Promote inbox assets into formal marketing-assets

Do not leave all value trapped in a one-off source folder. After building the JK asset, normalize reusable pieces into the shared marketing-assets layer.

Typical destinations:

• pain points → 02-pain-points/
• hooks → 01-hooks/
• angles → 06-angles/
• proof/source lines → 03-proof-points/
• CTA lines → 07-cta/

Rule:

• inbox/source asset = working note
• marketing-assets = durable shared substrate

At minimum, extract from the JK asset:

• reusable pain lines
• reusable hooks
• reusable angle lines
• source-backed proof lines

Read references/asset-promotion-guide.md before promoting shared assets.

8. Keep the content compliant

Always:

• attribute the source creator/video
• add original explanation and framing
• avoid copying long transcript passages
• avoid heavy reuse of original video/audio
• keep the result in commentary/education territory, not mirror-reposting

Read references/compliance-and-positioning.md before finalizing publishable outputs.

ClawLite bridge rules

Use soft bridges such as:

• “The concept is powerful. The usual blocker is setup friction.”
• “If you want to try this without the setup pain, start with ClawLite.”
• “This is the idea. ClawLite makes the first step easier.”

Avoid:

• overclaiming
• hijacking the creator’s work into a hard product ad
• turning every summary into aggressive CTA spam

Recommended output order

1. source note
2. beginner translation
3. short-video script
4. X thread
5. LinkedIn/Facebook post
6. short blog summary
7. ClawLite CTA bridge

Example use case

If given a source video like https://www.youtube.com/watch?v=fd4k16REDOU, produce:

• a summary note
• 3 key beginner takeaways
• a 45-second short script
• a ClawLite bridge angle
• a thread/post/blog content pack

NotebookLM automation layer

Use NotebookLM as the ingestion layer, not the final content layer. Its job is to help extract:

• transcript understanding
• summaries
• section structure
• notes and source context

Your real output should still be a ClawLite content pack.

When automating NotebookLM:

• screenshot before every action
• verify the modal/input target before typing
• avoid the sidebar search textarea
• re-dispatch input/change events when UI state does not update
• verify that the source was actually added before continuing

Read references/notebooklm-automation-guide.md before doing any NotebookLM UI automation.

Read next when needed

• references/source-note-template.md
• references/jk-marketing-asset-template.md
• references/angle-framework.md
• references/short-video-template.md
• references/content-pack-template.md
• references/asset-promotion-guide.md
• references/compliance-and-positioning.md
• references/notebooklm-automation-guide.md