Drop in any short-form video transcript. Get back five scored captions where Line 1 earns the watch and Line 2 earns the search ranking, ready to paste into Reels, TikTok, or Shorts.
← All Claude SkillsYou film the video. The caption is the part that decides whether anyone presses play. Caption Engine reads your transcript and hands back five caption options, each scored on curiosity, search signal, voice, and brevity, so you post the strong ones with confidence. Copy the two blocks below and give them to your Claude.
Step 1. Copy the skill block below.
Step 2. Open Claude and paste this instruction first, then paste the block underneath it:
Turn the block below into a Claude skill and install it. Create the skill file exactly as written. Then personalize it with smart defaults from what you know about me: my business, my voice, and my audience. Confirm when it is installed and tell me how to run it.
--- name: caption-engine description: Write scored short-form video captions for Instagram Reels, TikTok, and YouTube Shorts. Use this skill whenever you say "write captions," "caption this," "run the caption engine," "caption engine," "write the caption," or drop a video transcript (screenshot or text) and want captions for it. Also trigger when you mention Instagram captions, Reel captions, short-form captions, TikTok captions, or reference writing a description for a short-form video. If you drop a transcript screenshot and do not specify what you want, this is probably what you mean. This is the caption production system for all short-form video content. --- # Caption Engine — Short-Form Video Captions ## Purpose Turn any short-form video transcript into five scored caption options that stop the scroll and feed the algorithm. Every caption does two jobs: Line 1 earns the watch. Line 2 earns the search ranking. The video delivers the lesson. The caption never does. ## The Two-Line Template Every caption follows this structure. No exceptions. **Line 1: Curiosity Hook.** Create tension, contrast, or an open loop that earns the watch. Never state the lesson. Never give away the payoff. If someone can read this line and skip the video, it failed. **Line 2: Keyphrase-Rich Statement.** One natural sentence containing a searchable phrase the target audience would type into ChatGPT, Perplexity, or Google. This line expands the hook without closing the loop the video closes. That's it. Two lines. First line is for humans. Second line is for algorithms. ## The Process (Follow This Exactly) ### Step 1: Read the transcript If you drop a transcript screenshot, read it fully. If you paste text, read it fully. Do not start writing until you understand the complete message of the video. ### Step 2: Identify the core insight Strip away personality, metaphor, and delivery. What is the actual idea? Write it down internally in one sentence. This is NOT the caption. This is the thing the caption must never say outright. ### Step 3: Identify the keyphrase Ask: "What would someone type into ChatGPT, Perplexity, or Google to find this idea?" That's the keyphrase anchor for Line 2. Pick the most natural, searchable version. ### Step 4: Write five caption options Every option must follow the two-line template. Produce exactly five options. No more, no less. ### Step 5: Score every option Score each caption on four dimensions, 1-5 each. Display scores as bullet points beneath each caption. ## Scoring Criteria **Curiosity** — Does it create a reason to watch the video? Would someone stop scrolling? Does it withhold the lesson? If the caption delivers the payoff, it fails. (1-5) **AEO Signal** — Does Line 2 contain a natural keyphrase someone would actually search? Could an AI tool pull this caption as a relevant result for a real query? (1-5) **Voice** — Does it sound like you said it? Is it candid, direct, and grounded? Could this have come from anyone else? If yes, it fails. (1-5) **Brevity** — Is it two lines or less? Does every word earn its spot? Is there anything you could cut without losing impact? (1-5) **Minimum threshold to publish: 16/20.** ## Scoring Display Format Always display scores as bullet points directly beneath each caption option: ``` 1. [Caption Line 1]. [Caption Line 2]. - Curiosity: X - AEO: X - Voice: X - Brevity: X - **Total: X/20** ``` ## Caption Rules - Maximum two sentences. Period. - Never summarize the video in the caption. - Never state the lesson, takeaway, or conclusion. - Never use hashtags in the caption. - Line 1 must create an open loop, tension, or pattern interrupt. - Line 2 must contain at least one searchable keyphrase woven naturally into a sentence. - The caption should make someone NEED to watch the video to get the answer. - Curiosity comes from specificity and implication, not vagueness. "This changed everything" is vague. "The entrepreneurs pulling ahead all have one thing in common" creates pull. ## Voice Rules - Certain, grounded, warm, direct - Conversational. Like talking to a friend. - Write with presence and certainty. - Candid. The way you actually talk on camera. - NEVER use: delve, unlock, craft, foster, tapestry, embarking, unleash, unveil, realm, boasts, revolutionize, tenets - NEVER use semicolons, em dashes, contrastive framing, or tension-based setups - No corporate language. No AI filler phrases. - The caption should feel like you, not a content template. ## What the User Sees You drop a transcript (screenshot or text) and get back five caption options, each scored. Clean output, no preamble, no explanation of the process. Just the five options with scores. If you say "5 more" or "give me more options," produce five more following the same format. If you pick one, confirm it and move on. No upsell, no suggestions unless asked. ## Examples ### Example A — AI Practice Video Transcript summary: AI growth comes from hands-on daily practice, not passive learning. Sign up for tools and start creating. Core insight (internal only): AI growth is experience-driven. Keyphrase: "how to get better at AI" ``` 1. Everyone's talking about AI potential. Almost nobody is opening the tool. The fastest way to learn AI is to use it every single day. - Curiosity: 4 - AEO: 4 - Voice: 4 - Brevity: 5 - **Total: 17/20 — Publish.** ``` ### Example B — WordPress Disruption Video Transcript summary: WordPress may not survive 18 months. Saw a website built in 3 minutes by AI. Planning to make it his actual website. Core insight (internal only): AI website builders are replacing traditional platforms. Keyphrase: "WordPress" ``` 1. WordPress might not exist in 18 months. I saw why this week. - Curiosity: 5 - AEO: 4 - Voice: 5 - Brevity: 5 - **Total: 19/20 — Publish.** ``` ### Example C — Team Redundancy Video Transcript summary: One is none and two is one. If one team member leaves and everything stops, you have a liability. AI provides redundancy. Core insight (internal only): Single points of failure on teams are a liability AI can solve. Keyphrase: "team" / "AI for teams" ``` 1. Your social media person quits tomorrow. Does your content stop? That answer matters more than you think. - Curiosity: 5 - AEO: 3 - Voice: 5 - Brevity: 5 - **Total: 18/20 — Publish.** ``` ## Edge Cases - If the transcript is too short or unclear to extract a core insight, ask: "What's the main point of this video?" - If the video covers multiple ideas, pick the single strongest one for the caption. Captions do one job. - If you ask for longer captions, remember the template is two lines max and the system is built for short-form scroll-stopping. If you insist, follow your lead but flag that it breaks the optimization pattern. - If you ask for hashtags, do not include them. The system is designed without hashtags. If you insist, add them after the two lines but flag that it's outside the system. ## Output Note This skill produces the caption text only. Posting, scheduling, and platform formatting happen wherever you publish. The output maps directly to the caption/description field on Instagram, TikTok, or YouTube Shorts. No additional restructuring needed.
Claude reads the block, builds the skill file, and installs it. From then on, you just run the skill in any conversation. No folders to find, no code to write.
Every week you see what's working, build it live, and put it to work in your business.
Here's the shape of what this skill hands you when it runs. Copy it, use it, keep moving.
You don't need to hold any of this in your head. That's the skill's job. If you're curious what it's doing for you, here's the idea underneath it.
The video teaches the lesson. The caption has one job: get someone to press play, and be findable when the right person searches later. Caption Engine writes for both in two short lines and scores every option, so the only captions you ship are the strong ones. The watch and the search ranking, handled in two lines.
Every caption follows one structure. Line 1 is a curiosity hook. Line 2 is a keyphrase-rich statement. Two lines, every time, no exceptions.
The first line earns the watch by opening a loop the video closes. The second line earns the search ranking by carrying a phrase your audience would actually type in.
Every option is scored on Curiosity, AEO Signal, Voice, and Brevity, one to five on each. You see exactly why a caption is strong before you post it.
Anything under 16 out of 20 stays in the draft pile. The bar keeps weak captions off your feed automatically.
You always get five scored captions to choose from, so the best one for the moment is already in front of you.
It is the structure every caption follows. Line 1 is a curiosity hook that opens a loop and earns the watch without giving away the lesson. Line 2 is a keyphrase-rich statement that carries a phrase your audience would actually type into search. The first line is for humans, the second line is for algorithms, and the video delivers the payoff.
Every option is scored one to five on four dimensions: Curiosity, whether it earns the watch. AEO Signal, whether Line 2 carries a real searchable phrase. Voice, whether it sounds like you. Brevity, whether every word earns its spot. The four add up to a score out of 20, and the publish threshold is 16 out of 20.
Because the caption and the video have different jobs. If the caption states the takeaway, the viewer already has the answer and scrolls past. The caption creates the open loop and the video closes it, which is what earns the watch time the algorithm rewards.