YouTube Launch Kit

Turn a video topic, rough notes, or transcript into a complete YouTube launch package. Outputs two title variants, a keyword-optimized description with times...

Security Scan

The skill's requirements and instructions are coherent with its stated purpose (generating YouTube launch copy) and it makes no unexpected requests for credentials, installs, or external access.

Purpose & Capability

The name/description match the SKILL.md: it converts a topic/notes/transcript into titles, description, chapters, tags, pinned comment, and optional thumbnail/promo copy. There are no requested binaries, env vars, or installs that would be unrelated to this functionality.

Instruction Scope

Runtime instructions are narrowly scoped to parsing the provided topic/outline/transcript and producing the specified deliverables (titles, description, chapters, tags, pinned comment, etc.). The skill does not instruct the agent to read unrelated files, environment variables, or send data to external endpoints. It does recommend deriving timestamps from transcripts and using placeholders when timestamps aren't available, which is appropriate for the stated inputs.

Install Mechanism

No install spec or code files are present — the skill is instruction-only. This minimizes disk-write/execution risk and is proportionate to a copywriting/formatting tool.

Credentials

The skill requests no environment variables, credentials, or config paths. References to thumbnail generators (Midjourney/DALL·E) are only for prompt formatting and do not require keys or integrations, which is consistent with the README and SKILL.md.

Persistence & Privilege

Flags show normal defaults (always:false, user-invocable:true, model invocation enabled). The skill does not request persistent presence or modify other skills/settings.

Assessment

This skill is internally consistent and runs entirely as an instruction-only tool in your OpenClaw session. Before using it: 1) Do not paste sensitive or confidential content (transcripts may contain PII or private information) because the content will be processed by your model/session. 2) Generated tags/keywords are based on the skill's training — verify trending or competitive keywords with live tools like TubeBuddy or VidIQ for time-sensitive niches. 3) Thumbnail prompts referencing Midjourney/DALL·E are just text prompts; actually generating images requires using those external services and their credentials separately. 4) If you prefer the skill not be invoked autonomously by agents, adjust your agent's skill permissions or disable autonomous invocation for this skill in your settings. Overall, the skill appears safe and appropriate for the described use case.

Turn a video concept or recording into a complete launch package — titles, description, chapters, tags, pinned comment, discussion seed, and optional promo assets. One paste in, full launch out.

When to use this skill

Trigger this skill when the user:

• Pastes a video topic, rough notes, or outline and asks for a title, description, or tags
• Shares a transcript and wants launch copy generated from it
• Says "write my YouTube description," "optimize my title," "what tags should I use," "write my pinned comment," "write a first comment prompt," "write a discussion seed," "give me a thumbnail prompt," or "help me launch this video"
• Asks for a full YouTube launch package or upload checklist
• Wants promo tweets or social copy for a YouTube video

Do NOT trigger for:

• Channel strategy or long-term content planning — out of scope
• Video editing notes or scripting (pre-production writing, not launch copy)
• Analytics review or performance diagnosis — separate skill

---

Step 1 — Identify input state and collect what's missing

Determine which of three input states the user is in. Do not ask clarifying questions beyond what's listed below.

State A: Topic + notes/outline (primary)

User has a concept and rough structure but hasn't recorded yet.

Required:

1. Video topic (one sentence describing what the video covers)
2. Notes or outline — bullet points, rough section order, key points to hit, anything they've planned

Optional (use if provided, don't ask twice): 3. Target audience (e.g., "beginners," "freelance designers," "new parents") 4. Channel niche or name 5. Tone — educational, entertaining, motivational, tutorial-focused (default: educational/tutorial)

If notes are missing and only a topic is provided, proceed as State C and note the limitation.

State B: Transcript (fallback)

User has already recorded and has a transcript from auto-captions, Whisper, or another source.

Required: The transcript text (full or partial — work with what's provided)

Optional: Target audience, channel niche, tone

From the transcript: extract real chapter structure, pull exact phrases for keyword-dense copy, and use the creator's own language in the description and pinned comment.

State C: Topic only (edge case)

User has a topic but no notes and no transcript.

Proceed with title variants, tags, description intro paragraph, and a structural chapter scaffold. State explicitly once: "Chapter timestamps are structural placeholders — fill in the actual times after recording." Do not repeat this caveat.

---

Step 2 — Generate the launch package

Deliver all core deliverables in a single response, clearly labeled with H2 headers. Never skip a core deliverable. Offer optional deliverables at the end.

---

Title Variants (2)

Produce exactly two variants. Label them by intent, not by letter/number.

Curiosity-framed: Withhold one key detail or tease the outcome. Front-load the most compelling word or phrase. Drives browse traffic (suggested videos, homepage feed).

Clarity-framed: Direct, keyword-forward, searchable. Lead with the primary keyword. Drives search traffic.

Rules:

• Maximum 60 characters each (YouTube truncates beyond this in most surfaces)
• No manufactured urgency ("You WON'T believe...") unless the video genuinely delivers a surprise
• No all-caps words
• No trailing ellipsis in the clarity variant

---

Description

Structure in this exact order:

Opening paragraph (first 150–200 characters): Keyword-dense, hooks the viewer and states the video's premise. This is what appears above-the-fold in search results and what YouTube's algorithm indexes most heavily. Must contain the primary keyword naturally — do not stuff.

Body (2–3 sentences): Expand on what the viewer will get. Written in the creator's voice. Specific, not generic ("You'll learn the exact 3-step process I use" > "This video covers a lot of useful tips").

Timestamped chapters: Minimum 4 chapters. First chapter must be at 0:00. Labels must be content-specific — never use "Intro," "Part 1," "Outro," or "Conclusion" as standalone labels.

• State A/C inputs: Use structural timestamps in 0:00 / XX:XX placeholder format. The creator fills in actual times post-recording.
• State B (transcript): Derive timestamps from transcript structure. Use actual minute markers where inferable.

CTA section: One subscribe prompt + one resource link placeholder [LINK]. Keep brief — two lines maximum.

---

Tags (10)

Apply the tag formula exactly:

• 2 broad tags — high-volume, category-level (e.g., productivity, YouTube)
• 5 niche tags — specific to the topic and audience (e.g., notion for beginners, youtube growth tips)
• 3 long-tail tags — exact-phrase searches a viewer would type (e.g., how to organize your notion workspace, youtube tags for small channels)

First tag must exactly match or closely mirror the chosen title — this is YouTube's strongest tag-to-title relevance signal.

Present tags as a comma-separated list, ready to paste into the YouTube tag field.

---

Pinned Comment

50–80 words. Written in the creator's voice, intended to be pinned to the top of the comments.

Purpose: sticky reference — chapters, key resources, corrections, or links mentioned in the video. Not an engagement play; that's the discussion seed.

Format: lead with a chapter list or the most valuable resource, then link placeholders. Example structure:

⏱ Chapters are in the description — jump to whatever's most useful. 🔗 [Resource name]: [LINK] 🔗 [Resource name]: [LINK]

---

Discussion Seed

20–40 words. A question the creator posts as their first comment to seed early replies.

Rules:

• Must be specific to this video's content — never "What did you think?" or "Let me know in the comments!"
• Framed in first person from the creator's voice
• Asks something the viewer can answer from their own experience, not just by recapping the video
• Also called "first comment prompt" or "first pinned comment" in other frameworks — same output, different name

---

Optional Deliverables

After delivering the five core outputs above, offer both of the following once. Generate whichever ones the user requests (or both if they asked upfront):

Thumbnail Prompt (optional) A Midjourney or DALL-E compatible image prompt for the video thumbnail. Format:

• Subject description (person, object, or scene)
• Visual style (photo-realistic, illustrated, bold graphic, etc.)
• Text overlay suggestion (short phrase, ≤5 words, high contrast)
• Aspect ratio: 16:9

Promo Tweet Set (optional) 3 tweets, each ≤280 characters. Sequence:

1. Announcement — just published, what the video is about, link placeholder [YOUTUBE_LINK]
2. Hook — the most interesting insight, counterintuitive point, or moment from the video
3. CTA — subscriber-directed, drives watch + subscribe

Do not auto-generate optional deliverables unless asked. Offer them at the end of the core package.

---

Step 3 — SEO rules

Rules-based layer (apply to every output)

These signals are derived from known YouTube algorithm behavior — apply them without exception:

1. Description opening: First 100–150 characters do the most indexing work. Primary keyword must appear here, naturally integrated.
2. Title-tag alignment: First tag mirrors the winning title. No exceptions.
3. Tag formula: Enforced as written above — 2 broad, 5 niche, 3 long-tail. Padding beyond 10 provides no ranking benefit and dilutes relevance.
4. Chapter labels: YouTube uses chapter titles as secondary index signals. Descriptive labels outperform generic ones.
5. Title length: 60-character ceiling. Titles truncated in search results lose click-through rate.

LLM knowledge layer

Keyword selection and semantic variation draw on training knowledge of what searches well in a given niche. This works reliably for evergreen topics across most creator categories.

Honest ceiling — state this once in every session, do not repeat:

"Keyword recommendations reflect training knowledge, not live search data. For trending topics or highly competitive niches, validate tags with TubeBuddy or VidIQ after running this skill."

No external API key is required and none should be suggested as mandatory. TubeBuddy/VidIQ are validation tools, not prerequisites.

---

Step 4 — Style guardrails

• Never invent content. Do not include facts, statistics, claims, or timestamps not present in the input. If the notes don't mention a specific number, don't write one.
• No generic chapter labels. "Intro," "Main Content," "Part 2," "Conclusion" are banned as standalone labels.
• No "like and subscribe" in the pinned comment. Creators know to ask; including it reads as template filler.
• Match the creator's voice. If the notes use casual language, the description uses casual language. Don't formalize input that wasn't formal.
• Curiosity title must deliver. If the video input doesn't support the curiosity framing, use the clarity variant as the recommended title and note why.
• Description body is specific, not hype. "You'll learn the exact method" > "This video is packed with tips."

---

Step 5 — Follow-up offers

After delivering the package, offer once:

• "Want a version of either title tuned for a different audience or tone?"
• "Want the thumbnail prompt or promo tweet set?"
• "If you have a transcript, I can regenerate with accurate chapter timestamps."

Do not auto-regenerate unless asked.